Participants

Participants were N = 21 AMD patients (all Caucasian; 16 female, 5 male; age M = 83.5 years, SD = 7.3, range = 66 to 92). To be eligible to participate, patients had to: (a) be diagnosed by a qualified ophthalmologist as having AMD in both eyes and no other eye diseases (to ensure any vision-related problems were attributable specifically to AMD; note non-visually significant lens opacity was permitted); and (b) not have dementia (patients who disclosed a diagnosis of dementia during recruitment were not invited to participate, and all tested participants demonstrated normal levels of cognitive functioning during interview). Additionally, (c) patients had to report, on initial contact, experiencing difficulties seeing faces in their everyday life: while all patients with moderate and severe vision loss would be expected to experience face perception problems [11–15], early-stage AMD patients might not and, it is necessary for patients to report face perception problems to then interview them about the effects of those problems on social interactions and quality of life (i.e., patients not yet experiencing face problems would add no data concerning our major aims).

Participants were recruited until (a) we had covered a wide range of severity of vision loss from mild to legally blind (Table 1), and (b) saturation was reached in the qualitative interview results (i.e., no new experience types were being reported, the standard criterion for sufficient sample size in qualitative research, e.g., [23,24]). Patients were recruited through advertisement or individual approach from author JL, via: The Canberra Hospital Eye Clinic; a private ophthalmologist’s clinical rooms; local radio interview discussing the study; or letter from the Macular Disease Foundation Australia to AMD patients living in the Canberra region.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Individual patient details, with patients ordered by acuity (BCVA) in their best eye¹.

https://doi.org/10.1371/journal.pone.0209218.t001

Concerning demographics, the sample was generally middle-class and financially secure. For the 20 patients willing to answer financial questions, none disagreed with the statements “I have enough to pay my household bills” and “I have enough to pay for household repairs or help needed in the house”; only 4 disagreed with “I can afford to buy what I want”; 6 agreed with “I cannot afford to do things I would enjoy”. Regarding highest education level, 7 had a university qualification, 7 another tertiary qualification (e.g., certificate or apprenticeship), 5 secondary school and 2 primary school. Eighteen patients resided in their own house (8 still with a spouse), and 3 in assisted accommodation (e.g., nursing home). All reported regular contact and support from others (e.g., spouse, adult children, grandchildren, carers). Three participants were still driving.

Most patients were tested across three sessions, lasting up to 2 hours each. They were not paid, beyond reimbursement of travel to the university. The research was conducted in accordance with the Declaration of Helsinki. Ethics approval was obtained from the Human Research Ethics Committees of Australian Capital Territory (ACT) Health (protocol ETH.10.13.291) and Australian National University (protocol 2013/386). Participants' written consent was obtained, following explanation of the study and possible consequences; this included specific consent for the qualitative interviews to be audio recorded, and for publication of de-identified quotes.

Acuity, and criteria for mild, moderate and severe vision loss categories

Best Corrected Visual Acuity (BCVA) was measured by a qualified orthoptist using a retro-illuminated LogMAR chart mounted on a stand conforming to the ETDRS standard format [25]. Vision loss categories were defined using BCVA cut-off values from the International Statistical Classification of Diseases and Related Health Problems 10^th Revision [26]. Mild vision loss refers to BCVA poorer than 6/6 (normal vision), down to 6/18. Moderate refers to BCVA poorer than 6/18, down to 6/60. Severe refers to BCVA poorer than 6/60. To give concrete interpretation to the acuity values, in Australia, a standard driving licence requires BCVA better than 6/12, and 6/60 is legal blindness.

Patients were ranked (Table 1) and grouped based on best-eye visual acuity. This was on the grounds that AMD can affect the two eyes to different extents (in our sample, correlation between acuity in the two eyes was only r = .28), and it is functional acuity in the best eye which is likely to be the primary determiner of how well the patient can see faces in everyday life. This is because the brain preferentially attends to input from the eye providing the higher-quality input and tends to ignore input from an eye providing lower-resolution input; see evidence from amblyopia [27], or after laser surgery where the two eyes are given different corrections for close and far viewing [28]. S4 File provides: detailed information about both eyes (including BCVA, low contrast visual acuity LCVA, AMD type and stage); details of full vision assessments; and evidence that best-eye BCVA was indeed the most appropriate measure on which to rank patients' everyday vision ability.

Quantitative questionnaires: Everyday visual function, and psychological wellbeing

Overall level of everyday visual function was assessed using the National Eye Institute Visual Functioning Questionnaire (NEI-VFQ-25; Interviewer Administered Format plus appendix [21]). State (i.e., current) level of depression and anxiety were assessed using scales validated for elderly participants: the Geriatric Depression Scale-15; GDS-15 [29], and the Geriatric Anxiety Inventory; GAI [30]. AMD-related change in quality of life was assessed using the Macular Degeneration Quality of Life Questionnaire; MacDQoL [10], this instrument uses patients' ratings across multiple domains (e.g., ability to engage in hobbies, household tasks, travel outside the house, shopping, perform self-care) of how their life would change if they did not have AMD (but everything else in their life remained the same) multiplied by their rating of the importance of that domain to them. All questionnaires were administered verbally.

Qualitative interviews

Interviews were one-on-one. They were conducted at the ANU, the patient's place of residence, or (in a few cases) on the telephone. Each patient took part in one, or both, of two interviews, each lasting 30–45 minutes.

To explore the range of patient experiences, Interview 1 (all 21 patients) was semi-structured and open ended. The initial questions asked directly of all patients are listed in S5 File. S6 File gives examples of participants’ very different responses to a given initial question, and the corresponding variation in follow-up questions. Topics addressed were those described in the Introduction.

Interview 2 (19 patients, see Table 1) was primarily concerned with obtaining patient feedback on material for potential inclusion in our Faces and Social Life in AMD Information Sheet. We drafted a list of possible facts and statements, based on findings after testing most patients on Interview 1. We then asked patients in Interview 2 whether they did or did not endorse each fact/statement as useful to include in the Information Sheet, and to provide comments as needed (e.g., where they thought clarification or qualification was needed). In some cases, Interview 2 also revealed additional patient experiences, and/or included additional follow-up and clarification questions arising from their Interview 1 responses. Time delay between Interviews 1 and 2 ranged from same-day testing up to 2 years; vision assessment was repeated for the 4 patients with longer than 6 months delay (with none having moved to a more severe vision loss category across the delay; see Table A S4 File).

Interview data coding

Interviews were transcribed verbatim, combined across Interviews 1 and 2, and entered into NVivo software (QSR International Pty Ltd, Version 10, 2015) to assist with data collation. Patient experiences from the transcripts were coded into themes [31], using a mix of a bottom-up (inductive) and top-down (theoretical) approaches. For bottom-up analysis, three authors independently read 6 interviews (3 from Interview 1 and 3 from Interview 2) from patients spread across the three AMD severity levels, and extracted emergent themes relevant to the present research questions (i.e., content related to face perception and/or its effect on social interactions and quality of life); coding strategies between authors were reviewed, and themes chosen were based on consensus negotiation [32]. Themes were also redefined (e.g., two sub-themes combined), and additional themes were developed, in a top-down manner, to ensure adequate coverage of all the specific topics we wished to address (e.g., emotion perception; technology preferences), and to allow comparison to previous findings in the literature (e.g., whether AMD patients experience the same types of identity-related social-interaction problems reported in prosopagnosia). JL then coded the interview transcripts from each patient into the final themes, including whether the patient had experienced that type of phenomenon or not, together with the piece/s of quoted text relevant to that experience. EM cross-checked the coding, with discrepancies resolved via negotiation. In addition to the initial coding of full interview transcripts to themes, multiple text search queries were conducted for each theme to avoid missing any data.

Quantitative measures of visual function and psychological wellbeing

Table 2 presents sample-descriptive results for quantitative scales, including across our full patient sample, and subgroup means for patients with mild, moderate and severe vision loss (ICD-10 criteria [26] as used in Table 1). Table 2 also shows correlations with acuity (best-eye BCVA from Table 1, converted to logMAR; note higher logMAR scores indicate poorer vision).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Patient results and comparison values for quantitative questionnaires.

https://doi.org/10.1371/journal.pone.0209218.t002

As expected, Table 2 demonstrates impairment in self-reported everyday visual function. All individual patients reported meaningful everyday impairment on the NEI-FVQ (highest score = 69.8 where 100 is no impairment). Self-reported NEI-VFQ function correlated significantly with objective vision level, with function worsening with worsening acuity (significant negative correlation with BCVA). We also found impairments specifically on the two NEI-VFQ items that are relevant to face perception (Question A6 ‘because of your eyesight, how much difficulty do you have recognizing people you know from across a room?’; and Question 11 ‘because of your eyesight, how much difficulty do you have seeing how people react to things you say?’). Mean scores for both items were well below the no impairment level, and every patient indicated impairment on the identity and/or the expression question (i.e., the highest scores of 100 indicated in the range data in Table 2 came from different patients). Both face-relevant items showed correlations with objective acuity that were in the predicted direction, significantly so in the case of the face-identity-related item (Question A6).

For psychological wellbeing measures, Table 2 shows worsening acuity correlated significantly with increasing anxiety. Depression did not correlate linearly with worsening acuity (a finding consistent with evidence of psychological adjustment to chronic disease [33]). Results for the MacDQoL showed a sizeable AMD-associated reduction in quality of life on average (i.e., mean score of –3.9 where 0 is no impact), with a close-to-significant correlation between worsening acuity and greater AMD-associated reduction. Note there are no face-related-item data provided for the MacDQoL in Table 2 because the measure includes no items from the face domain.

Qualitative experience of AMD patients

To illustrate the range of experiences patients reported in the interviews, we use a mix of quotes (from which irrelevant information has been removed, and any names changed) and tables containing the percentage of patients reporting certain experiences. These percentages are minimum values in many cases (marked with a '+' in Tables 3–7); this is because, in a semi-structured interview procedure, not all patients are necessarily asked directly about all experience types, meaning additional participants within our sample may have endorsed the experience if explicitly asked about it. Our reason for reporting concrete numbers at all—which is unusual in a qualitative interview study—is to provide information on whether a given difficulty was reported, say, only by patients with severe vision loss, or also reported by patients with mild vision loss.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Difficulties seeing faces in AMD, how faces appear to patients, and problems with face identity and expression recognition.

https://doi.org/10.1371/journal.pone.0209218.t003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Alternative strategies that AMD patients try to use, and their effectiveness.

https://doi.org/10.1371/journal.pone.0209218.t004

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Difficulties with, and changes to, social interactions arising from poor face perception.

https://doi.org/10.1371/journal.pone.0209218.t005

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 6. Other face problems: Eye gaze and facial speech.

https://doi.org/10.1371/journal.pone.0209218.t006

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 7. Impact of reduced face perception on social life, confidence, and quality of life; plus lack of understanding by other people.

https://doi.org/10.1371/journal.pone.0209218.t007

Difficulties seeing faces, facial appearance, and variability in face perception

Table 3 collates reports of difficulties seeing faces. Results show that all patients, regardless of their residual visual acuity, reported their vision loss had made it harder to see faces. Problems were described as particularly acute at longer distances (e.g., across a room; also see the NEI-VFQ results in Table 2), but nearly half of patients reported having problems seeing faces clearly even at conversational distances (1–2 metres), including three mild patients.

Concerning how faces appear visually to people with AMD, Table 3 shows three-quarters of patients spontaneously mentioned one or more ways in which faces no longer looked normal. The most common aspect mentioned was that faces appeared blurred (or equivalent terms such as ‘unclear’, or ‘low-definition’). This blur meant that patients could not always see internal features clearly. For example, one patient described the interviewer, sitting less than 2 metres away, as having ‘two holes for the eyes’ (P1; mild). Additionally, nearly a third of patients mentioned seeing shape distortions and missing parts in the face. The nature of these varied: one patient said ‘The distortion is quite bad … on one side the mouth goes up and the eyes keep disappearing … or looks blurred and moving a bit’ (P2; mild); another said the ‘features are kind of deformed, jumbled … it’s as if the face were on a piece of sheeting or something and somebody grabbed it from behind and pulled it like that [simulating a sheet being grabbed] and it just went all scrunched up’ (P9; moderate); another said ‘I can see the right hand side of you … not the left’ (P19; severe). One third of patients also reported other general visual disturbances (e.g., black flecks, lights and floaters) that would impact on the appearance of faces. Three patients mentioned seeing a black blob in the centre of their vision (a common illustration of the supposed perceptual effects of AMD; see Discussion), while one patient (P18; severe) specifically said they did not experience a black blob in the centre, instead describing their experience as like ‘looking through a screen’ or ‘looking through black tulle’.

Table 3 also shows that many patients mentioned variability and inconsistency in how well they could see faces. Lighting was reported as a relevant factor by most (e.g., one example was that it was harder to recognise faces with the light behind them). Eleven patients said they prefer strong lighting, with faces harder to see in lower light levels. Three said the opposite, namely that they are light sensitive and prefer low lighting. Two said their light preference varies, i.e., sometimes they require strong light and other times they are light sensitive. One of these latter patients commented ‘This is one of my husband’s big bug bears, because he just can’t understand why one minute I want light and the next minute I don’t’ (P13; moderate). Some patients identified other factors associated with variability in how well they can see faces, including time of day (e.g., improvement as the day goes on), and treatment phase (i.e., pre/post injection if being treated with ranibizumab for Wet AMD). One patient said: ‘Sometimes I can see, sometimes I can’t’ (P19; severe).

Difficulties with face identity and expression recognition

Table 3 shows that all patients reported their difficulties seeing faces resulted in problems recognising both face identity (who other people are) and facial expression. Importantly, the problems were not limited to those with moderate and severe vision loss, but also occurred in mild vision loss.

For facial identity, both false negatives and positives were common. Almost all patients had experienced problems recognising people they know (false negatives). This included reports of failing to recognise good friends and close family members. Problems occurred even with mild vision loss, for example ‘I have had it happen, it’s very embarrassing … the other day I didn’t even recognise my son … within a yard or two of me and I didn’t recognise him, he said “Mum, it’s David!”‘ (P2; mild). In general, patients with more severe vision loss reported such failures occurring more frequently. When asked ‘Do you find that you fail to recognise people you know?’, responses included ‘Oh all the time (P18; severe)’, and ‘I would pass people by in the street that I know very well. It’s very embarrassing … they’ve said to me “hey Mary how are you?” … I would not have a clue … I don’t mean to be rude but I just can’t see them’ (P21; severe).

Many patients also reported falsely ‘recognising’ people they did not actually know (false positives), such as approaching a person to say hello thinking they were a friend, only to find they were a stranger. When asked ‘Have you said hello to someone thinking it was someone you knew, and it actually wasn’t?’ example responses included ‘Yes and it’s not them at all. It’s someone totally different, yes, that becomes a bit embarrassing’ (P19; severe), and ‘Yes [laughs] … Someone I knew very well, I went up and started having a conversation with them and they looked at me blankly. And you know when I was closer: “yeah, you’re not who I thought you were”. I apologised to them, but what they thought, I don’t know’ (P14; moderate). These experiences of false positives and negatives in everyday life closely match quotes describing identity-related failures in prosopagnosia [17,35].

Turning to facial expressions, comments suggested expression perception was even more severely affected by AMD, earlier in the progression of the disease, than identity recognition. When asked: Has AMD impacted your ability to see a person’s facial expressions? example responses were: ‘Yes, I think that was one of the first things that went, not actually see the expression’ (P5; mild), and ‘As far as expressions go, that’s something that’s gone’ (P16; moderate), and ‘Well you don’t get facial expressions with this disease’ (P18; severe).

Alternative non-face-based strategies, and their effectiveness

Problems identifying faces and recognising facial expressions would not have serious implications for social interactions if AMD patients were able to use other strategies to successfully recognise people and their emotions. However, this was commonly not the case.

Table 4 lists various alternative strategies that patients reported trying to use (not necessarily successfully). For identity recognition, the most common visual strategy mentioned was using body shape/size followed by walk/gait, hair/hairstyle and clothing. Patients also reported two nonvisual strategies, voice recognition and context. Context was identified as both a help ‘When it’s a normal meeting it’s not so bad because most of them sit on the same tables’ (P20; severe) and a hindrance (when people are seen out of their usual context or when the patient was expecting someone else). Most patients reported using multiple strategies simultaneously, for example ‘I look at the way people are walking … mainly their gait and their general appearance … maybe for the women I look at their hair … their hairstyles … their size and behaviour … and then of course if they speak it’s voice recognition’ (P18; severe). These identity-related strategies were identical to those that prosopagnosics report trying to use [17,35,36]. For expression, the two main strategies reported for trying to understand other people's feelings and emotions were using body language (e.g., a sad or angry posture) and voice: ‘The tone of voice gives them away. Mostly it’s reflected in people’s voices whether they are in a happy mood or a grumpy mood’ (P11; moderate). A strategy used by many patients, relevant to both identity and expression, was proximity, i.e., moving closer to others, or waiting for others to approach to try to improve the clarity of the face. Overall, these alternative strategies involved either looking at visual information that survives low resolution vision relatively well (e.g., because the body is larger than the face, or because determining hair colour and length requires only coarse spatial information), or using nonvisual information (auditory cues, context).

Importantly, patients reported that the non-face-based strategies they tried were often ineffective. Table 4 shows half the patients reported their suite of strategies failed: for example, ‘I can make some terrible mistakes … [my strategies] help, they are certainly not fool-proof’ (P11; moderate), and ‘I’m not sure that [my strategies] are very effective at all’ (P1; mild). Moreover, even when patients initially described their strategies as ‘effective’, further interview responses revealed patients generally meant the strategies worked ‘most of the time’ or ‘in some contexts’ (e.g., for close family members). Patients also reported factors that can impair the effectiveness of their strategies; these factors included crowds, how often they see the person (i.e., level of familiarity), and the fact that some strategies are unreliable due to changes in the environment. Environmental changes included a participant who misidentified her own husband because he had recently lost weight and his body shape had changed (P13; moderate), and another who said ‘One of the ladies at church had long curly hair and she got it all chopped off and I didn’t have a clue who she was until she spoke’ (P14; moderate). Finally, the effectiveness of alternative strategies appears to decrease (Table 4) as AMD severity level increased. This may be because visual cues that survive mild loss of visual acuity well (e.g., body shape) become too blurred to be useful in moderate-to-severe AMD.

The failure of AMD patients' alternative strategies meant they often reported being reliant on other people for assistance. In total, 86% of patients reported others helped sometimes, by naming the person aloud (e.g., ‘here comes Bill’ or ‘James is sitting at the back of the room with his wife’) or describing emotions (e.g., ‘Mary looks sad today’ or ‘Jan was smiling when she said that Mum’). Additionally, however, 7 patients reported that they would appreciate more help of this kind.

Difficulties with, and changes to, social interactions

Table 5 collates patient responses concerning the ways in which their face perception problems alter their immediate social interactions.

Longer-term impact on patients’ social life, confidence, and quality of life

Table 7 summarises information about the longer-term effects of poor face perception on patients' social life, confidence, and quality of life. Negative impacts were common.

Resilience

In addition to the difficulties described above, it should be noted that most patients revealed examples of resilience, some degree of coming to terms with their face problems, and/or determination to fight against social isolation. Quotes include: ‘[not being able to see facial expressions] used to make me feel upset … it was as if I’d lost something, lost a person sort of, but I’ve kind of got used to it now … You either get up and go or you sit in your chair and die and I think I’d rather get up and go’ (P5; mild); ‘I think it [social withdrawal] could happen to people but I don’t let it happen to me … I imagine that AMD could [make me more passive] and maybe it will get me that way eventually, at the moment I am still fighting it’ (P2; mild); ‘I’ve got on with life the best I can’ (P14; moderate); ‘Unless you sort of make an effort [socially], you could have a very miserable life, and I’m just not prepared to have a very miserable life’ (P13; moderate).

'Other people don't understand'

Table 7 shows that most patients felt others did not understand how AMD affects their vision, and how hard it was for them to see faces clearly. In some cases, this occurred even when the other person was well aware of the AMD: ‘They know what I am like, at that minute they just forget’ (P16; moderate), and ‘They don’t realise [my vision’s] deteriorated yet’ (P8; mild). In some cases, it reflected the other person having difficulty understanding that AMD can affect fine vision as needed for face recognition, but without externally-visible damage to the eyes, and without impairing coarse vision as needed to walk around. For example: ‘A couple of people have specifically said to me, in appearance, you don’t look as though you are having problems with your vision, your eyes look perfectly normal’ (P5; mild), and ‘I have lots of friends and family even who say to me “well how can you see that and you can’t see something else?”‘ (P13; moderate).

This lack of understanding resulted in more than a third of patients raising concern that others think they are 'faking it'. For example: ‘I think she [my carer] thought I was just putting it on a bit you know’ (P8; mild), and ‘He [my son] had been a bit doubtful I think about it, but that [my failure to recognise him] has quite convinced him now that faces are distorted for me’ (P2; mild). This problem was not limited to patients with only mild vision loss, where others might perhaps be expected to be least appreciative of AMD difficulties (e.g., because the patient can still navigate well around the environment). Examples from moderate vision loss included: ‘sometimes you sort of think they’re doubting that you even have a problem’ (P13; moderate); and ‘I feel like a fraud [when I need to ask for help]’ (P9; moderate). Even patients who were legally blind worried that others didn't believe they had a real problem seeing faces: ‘I do think people think you are faking it a bit’ (P18; severe).

Relative importance to patients of faces versus other aspects of vision loss, and of facial identity versus expression

Three questions addressed how important patients perceived their face problems to be, including relative to the many other visual domains affected by AMD (e.g., ability to read, drive, or cook). At the very beginning of Interview 1, we asked an open-ended question—‘Which areas or tasks have been made harder because of your AMD?’—to record how many patients would spontaneously mention face perception difficulties as amongst their most important everyday problems: 38% did so (3 mild, 4 moderate, 1 severe). This compared to 81% mentioning reading and 62% hobbies (e.g., knitting, sewing, crosswords and writing) as the two most common topics raised, and to rather smaller percentages for some domains included in the MacDQoL [10] (e.g., only 5% for enjoying meals and 14% for shopping; see S7 File for details). Later we asked directly for ratings of importance. In the identity section of Interview 1, in response to ‘How important is seeing other people's faces to you?’ 15 patients indicated High/Very High, 5 said Medium, and only 1 said Low. In the expression section, in response to ‘How important is it for you to be able to see another person's facial expressions?’, 16 patients indicated High/Very High, 4 said Medium, and only 1 said Low (a different participant from the person who chose Low for identity). Together, results of these three questions indicate that face perception is of high importance to most AMD patients, and also of no lesser importance than several domains currently included in the MacDQoL [10].

We also asked patients which was more important to them, recognising facial identity or recognising facial expressions: 67% said it is more important to be able to recognise who a person is (i.e., identity); 24% reported the two are equally important; and 10% said it is more important to see a person’s facial expressions.

Patients’ views on possible technological help

Our results so far indicate that technology that could improve patients' face perception and recognition would have the potential to improve their social interactions and quality of life. Our qualitative interviews thus addressed patients' willingness to try various types of future technological assistance. Most patients (71%) said they would be willing to use computer facial recognition software that could say aloud the name of other people in the environment. While a commercial product offering this service is available (for 100 learned faces [40]), note the named-by-computer approach has some disadvantages compared to improving a participant's own perception: naming aloud may interfere with hearing an approaching person speak, and is also only useful only for identity recognition (a running verbal commentary on dynamic aspects of faces like expressions or eye gaze would not be suitable for patients). We thus also asked about image enhancement, which has the potential to improve perception of all aspects of faces. We explained that, in image enhancement, patients would view faces on computers, tablets, or smart glasses [41,42] with the faces digitally altered in ways that have been reported to improve low-resolution face perception (e.g., making the face larger [15,43]; increasing the contrast of medium and high spatial frequencies [44,45]; or caricaturing the face, i.e., exaggerating its appearance away from the average to make identification easier [46–48]). Participants were on the whole very positive about trying image enhancement technology if and when it became available, particularly on TV and/or computer screens (90% support; e.g., one patient mentioned this would be useful when skyping his family), and via smart glasses (90% support). Only one patient (P16) was not interested in any type of electronic device, saying technology ‘left me behind’. Several patients noted that hand-held screens (e.g., phones, tablet computers) would not be valuable because they needed their hands available for balance, carrying things, and in some patients for using a mobility walker; thus, a hands-free option such as smart glasses was preferred. Several participants highlighted potential concerns regarding the weight of smart glasses on the face, and the likely expense of the technology (even in our largely middle-class sample).

Development of information sheet, conversation starter and brochure

In Interview 2, all but one participant asked indicated that an information sheet concerning the effects of AMD on face perception and social interactions would be useful. Responses from Interview 1 were used to draft potential points for inclusion in the information sheet (see Method). The final version contained concepts that were directly endorsed by patients for inclusion (for details, see S8 File), plus some additional points widely raised by patients (Tables 3–7) which we had not included in the draft but emerged as important themes once we had the full data set to analyse. Feedback on a final draft was also obtained from the Macular Disease Foundation Australia (via their Research Officer and Medical Affairs Manager), concerning appropriate formatting and language for macular degeneration patients.

Our final community resource materials are in S1–S3 Files. The content of the Faces and Social Life in AMD Information Sheet highlights the core issues arising from the patient interviews: how faces might look to people with AMD; how early in disease progression problems could potentially arise; the types of social problems patients might experience that result from difficulty seeing faces; that individual experiences can be highly variable; and that the patient's experiences are genuine and normal for AMD. The sheet also aims to help others around the patient (e.g., family members, friends, carers, nursing home staff) appreciate compassionately why a patient may sometimes appear to behave in a socially odd manner (including appearing rude or unfriendly); to understand why they may have become more passive or disengaged in social settings; and to provide suggestions for how others might be able to help in concrete terms (e.g., by naming people as they approach). To give greatest accessibility to vision impaired patients (as advised by Macular Disease Foundation Australia), the information sheet has been prepared in large font, all plain text (no italics), and maximum contrast (plain black text on white background).

Our accompanying Conversation Starter has questions for others to ask the individual person living with AMD. These are aimed at enabling AMD patients to describe in detail to family, friends and carers (e.g., nursing home staff) their own individual experiences in seeing faces, the impact this is having on their social interactions and social life, and what they would like other people to do to help. Finally, we provide a tri-fold 1-page brochure that can be given to patients at vision clinics, or picked up by family members, containing directions to finding the information sheet and conversation-starter online; the brochure and information sheet are provided in both A4 and US-letter paper sizes.

Implications for the current Quality of Life instrument for macular degeneration (MacDQoL)

A key implication of our findings is that problems with face perception in AMD contribute to social difficulties and reduced quality of life, over and above all of the other functional visual problems that occur in AMD (i.e., inability to drive, engage in hobbies, remain independent, etc.). This argues the MacDQoL [10] would benefit from adding a specific item about face recognition.

The MacDQoL is the only quantitative questionnaire designed specifically to measure change in Quality of Life due to macular degeneration. It is widely used [49], is available in 14 languages (https://www.healthpsychologyresearch.com/information/currently-available-translated-questionnaires/MacDQoL-macular-disease-dependent-quality), has good reliability, and has been validated overall (e.g., scores correlate with level of vision loss; [22,50]). However, its 23 items include no questions addressing the domain of face perception. Additionally, while it does include questions relevant to social relationships with others ('closest personal relationships'; 'family life'; 'friendships and social life'), these questions do not mention faces, and patient responses could equally be related to other non-face difficulties: for example, a patient's personal relationship with their spouse may be negatively impacted by the fact they can no longer contribute to household tasks; or a patient's friendships and social life may be reduced due to loss of driver's licence. Wording for a face-domain question for the MacDQoL could be along the general lines of asking the patient how much better they could see other people's faces (e.g., recognising who they are, or what they are feeling) if they did not have macular degeneration, and then following up with the usual MacDQoL question structure by asking participants how important seeing other people's faces is to them. (Unfortunately, we cannot suggest here precise wording that would match the MacDQoL format, because the questionnaire and its precise format is copyright).

Our present study also revealed a problem with one of the existing items on the MacDQoL [10]. Q15 asks patients about how much better other people would react to them if they did not have macular degeneration (with choice between 5 "amount" options ranging from very much better to worse). However, consistent with evidence that AMD patients often cannot see how others react to them, due to poor facial expression perception, 7 of our 21 patients said they could not answer this question—indeed, one said "How do I know?". The inability to see how others react is a well-established aspect of vision loss of multiple types (e.g., as reflected in the NEI-VFQ’s inclusion of the item 'Because of your eyesight, how much difficulty do you have seeing how people react to things you say?'), further arguing that asking patients about how much macular degeneration has affected how others react is not a useful wording.

In sum, our present results argue the MacDQoL could be improved in validity by (a) adding a question about quality of life in the domain of seeing faces, and (b) removing or rewording Q15.

Improving community and patient information about AMD

Concerning community information about AMD, our interview results have revealed two key findings: poor face perception in AMD is qualitatively related to many difficulties in social interaction; and patients commonly feel that others don't understand these difficulties. This argues that one way to improve patients' quality of life is to improve community understanding of face-related social difficulties in AMD.

Our Faces and Social Life in AMD Information Sheet (and accompanying conversation stater and brochure) provides a much higher level of detail concerning face perception and social interaction difficulties than previously-available public material. For example, beyond noting that faces can be hard to see, macular disease organisations including the Macular Society (https://www.macularsociety.org/), the American Macular Degeneration Foundation (https://www.macular.org/), and the Macular Disease Foundation Australia (https://www.mdfoundation.com.au) have previously given no information about how this can affect social interactions. Vision Australia, which deals with all vision disorders including total blindness (http://www.visionaustralia.org), has an information page briefly explaining that vision loss can result in difficulties with social interactions due to poor face perception. This page mentions problems seeing facial expressions and maintaining eye contact, but does not mention facial identity. As the present results show, problems with identity recognition can have very important negative impacts on social interactions and social life (also see [17,35] in prosopagnosia). Additionally, the Vision Australia page does not refer to macular degeneration specifically, nor explain that AMD patients might sometimes experience face-related social interaction problems even when their vision is still otherwise quite good (e.g., good enough to drive).

Our hope is that our new community materials will help others around the patient (e.g., family members, friends, carers, nursing home staff) appreciate compassionately why a person living with AMD may sometimes appear to behave in a socially odd or changed manner (including appearing rude, unfriendly, or unusually passive), and to help to assist with maintaining social engagement in practical ways (e.g., by naming people as they approach).

Qualitative results provide a basis for future development of a quantitative questionnaire

An important limitation of the present study is that we have not aimed to examine the frequency or severity with which different types of face perception and social interaction difficulties occur. There are many research questions of scientific interest that can be addressed only with access to such quantitative information. For example, these include: (a) in multiple regression, how much of the decreased quality of life in AMD is uniquely related to face perception and social interaction problems over and above, say, the contributions from other aspects of vision loss (e.g., loss of ability to read or drive) or general age-related difficulties (e.g., other health problems, death of spouse or old friends; (b) whether a potential intervention (e.g., psychological treatments for social anxiety; advice to disclose AMD-related face problems to others; technology to improve face perception) produces a significant improvement in social interaction and quality of life by comparing pre- versus post-intervention scores; (c) whether certain face perception or social interaction difficulties might be more severe in certain ophthalmological states (e.g., wet versus dry AMD) or perceptual states (e.g., the patient experiences blur-plus-distortions in the face, or only blur); and (d) the extent to which AMD patients have insight into their precise level of deficit (i.e., by correlating quantitative self-report of everyday face problems with lab-based measures of face ability), noting that insight is limited in normal vision [51,52] but perhaps may be better in patients given the potential benefit of internal comparison to their earlier abilities before AMD.

Development of a validated quantitative questionnaire to measure face-related social interaction difficulties is beyond the scope of the present study. Importantly, however, the qualitative analysis presented provides a crucial first step towards this end, by providing evidence on the types of face perception and psychosocial difficulties experienced by AMD patients. This detailed understanding of the types of experiences that can occur is an essential step towards developing a valid quantitative measure. Research that creates quantitative self-report measures without prior qualitative understanding can result in validity problems. For example, in the 8-item self-report questionnaire [15] developed to supposedly assess AMD patients' insight into their level of face recognition difficulties, four items do not actually ask simply about perceived disability in face perception/recognition per se: two ask about emotional responses to mistakes; and two ask about alternative strategies and wrongly assume that the only alternative strategy is voice.

Our qualitative results indicate that important domains to assess quantitatively, depending on the specific interests of the researcher, could include: frequency with which everyday face recognition problems are experienced (including identity, expression, eye gaze, and facial speech); degree of success or failure of alternative strategies to recognise people and their emotions other than by their face; frequency and severity of various types of social interaction difficulties related to impaired face perception; severity of negative emotional response to making errors (noting our participants varied on this, with some highly embarrassed and trying to avoid errors at all costs, and others more inclined to shrug off many of their mistakes); and severity of face-related impacts on social functioning, confidence, and quality of life. Potentially, items for a formal questionnaire could be based on the experience categories in Tables 3–7 and/or include rating-response versions of questions included in our Conversation Starter (S1 File).

To fully develop a quantitative self-report questionnaire from the present results would require large-sample testing to determine psychometric properties for any proposed instrument, including reliability (e.g., Cronbach's alpha, test-retest), factor structure, and convergent and divergent validity. For example, there may be separate factors for severity of face perception problems and severity of negative emotional responses to those problems; if so, we would expect face perception to correlate most strongly with acuity and lab-based face tests, while emotional response might correlate less with acuity and more with personality attributes.

How widespread might social-interaction problems be in patients with only mild vision loss?

A potentially surprising finding from the present study was the consistent reports of face-related social interaction difficulties even in AMD patients with mild vision loss. While it may of course be the case that these occur quantitatively less frequently than in patients with more severe vision loss (an issue which requires development of a quantitative questionnaire to evaluate), qualitatively the problems were the same as in the moderate and severe vision loss categories. In terms of face identity failures, perhaps the most striking was our second mildest patient, who despite having visual acuity of 6/9.5 (best-eye BCVA), reported recently failing to recognise her own son standing next to her. Concerning expressions, a patient with acuity 6/15 said expressions were ‘one of the first things that went’. In terms of social interaction effects of poor face perception, even our best-acuity patient (acuity 6/7.5) said ‘you’re not with the rest of the crowd … you’re not getting out of a conversation perhaps what you would normally get out’.

Noting that we specifically recruited AMD patients who reported on initial phone contact that they experienced face recognition problems in everyday life, an open question is how widely spread face-related social interaction difficulties will be amongst mild vision loss patients. Predictions vary depending on the possible cause.

First, predicting that social problems in mild-vision-loss AMD patients would be common, is a potential role for low-contrast visual acuity. Low- and high-contrast acuity are highly correlated (r = .93 in our sample) but absolute acuity performance is always poorer for low contrast than high contrast stimuli (see lower LCVA than BCVA for patients in Table 1; also for normal-vision observers [53,54]. Acuity for static, high contrast letter stimuli is not fully reflective of performance in real-world viewing conditions [55,56]. Faces, in particular, are dynamic stimuli seen in changing conditions of lighting and contrast, and also intrinsically contain much low contrast information, such as the shape of the boundary between the nose and the cheeks for face identity, or the presence of frown lines in the forehead for expression. Thus, the BCVA measure, and the ICD-10 [26] categories based on it, is likely to underestimate the absolute degree of functional vision loss relevant to perceiving faces. The greater absolute vision loss implied by LCVA suggests social interaction difficulties would likely be widespread in other 'mild' patients.

Alternatively, predicting that social problems in mild-vision-loss AMD patients may be more restricted is the idea that there may be something special about mild patients who experience face perception problems often enough to noticeably impact their social interactions. One hypothesis is that there is something specific about the pattern of retinal damage in our patients that account for face perception difficulties severe enough to cause social interaction problems (e.g., perhaps such patients tend to have their relatively-well-preserved acuity supported by a single small region of preserved retina within the fovea, rather than by peripheral vision); note that, while we have detailed retinal data for our patients (S4 File), we cannot yet evaluate this idea empirically because we did not recruit mild-vision-loss patients without any face problems for the present study. Another hypothesis is that there might be an effect of the other eye for our patients. While other-eye acuity did not predict functional vision levels across our full sample (e.g., on the NEI-VFQ; see S4 File), it is of some note that of our 7 mild vision loss patients, 6 had severe vision loss in the other eye (i.e., BCVA worse than 6/60; see Table A S4 File). Thus, it cannot be ruled out, for example, that while typically patients preferentially use input from their best eye and ignore input from the poorer eye [27,28], our particular mild patients might be more likely than average to experience breakthroughs into attention from the poorer eye, and that these impair ability to perceive faces (e.g., if the severely-impaired other eye sends input suggesting unexpected movement in the face, such as a distortion or a part disappearing or reappearing, noting that movement attracts attention).

Are negative consequences of poor face identity recognition due to having a disease not known to the general public?

In addition to the major issues arising from our study discussed above, a number of brief points arise concerning a variety of topics.

Yardley et al. [17] hypothesised that, in prosopagnosia, low public awareness of the disorder at the time may have been a major contributor to feelings of embarrassment, guilt, and failure. Awareness of AMD in Australia is very high (in 2011, 80% of people aged over 15, and 92% of people aged 50+ years were aware of macular degeneration, and 73% understood it is a disease of the eyes [57]). Despite this, we still found feelings of embarrassment, guilt, and concerns about being perceived as a fraud (also see [9]). It thus seems that the critical variable here is not public awareness of the medical condition per se. Instead, it may be that AMD is an invisible condition (noting experiences of being treated as fraudulent or exaggerating are reported in other medical conditions not easily visible to others, e.g., endometriosis [58,59]), and/or that others do not have sufficient information to understand the specific ways that AMD actually affects vision (i.e., that it impairs fine vision tasks far more than coarse-vision tasks). This argues improving the patient experience requires increasing understanding of the detailed symptoms of AMD, rather than merely its existence.

How faces appear to patients with AMD: Blurred, distorted, variable, and often not the central black blob of traditional illustrations

Surprisingly, no previous study seems to have asked AMD patients in any detail what faces look like to them. Taylor et al. [60] recently provided the first detailed self-reports about visual appearance in (dry) AMD, covering visual experience in general rather than specifically faces. Across 29 patients with geographic atrophy, the most highly reported descriptors were blur (45%), missing parts of the image (34%) and distortions (24%). For faces, our present results agree in broad terms, with blur the most common phenomenon reported (71%) and reports of distortions and/or missing parts of the face the next most common (29%; patients with dry and wet AMD). Also in agreement with Taylor et al. [60], we found considerable heterogeneity in reported appearance across different patients.

These results are important because they indicate that the most common illustration of how the world is supposed to appear to patients with AMD is inappropriate (e.g., National Eye Institute, NEI, https://nei.nih.gov/health/examples; also [61]). The NEI illustration would often be viewed by newly-diagnosed patients wondering what to expect as their AMD progresses, and shows a black or grey blob completely hiding faces located at central vision. However, in [60], only 2 of 21 patients said this type of image reflected a good depiction of their vision. Here, only 3 of our 21 patients spontaneously mentioned seeing a blob in their central vision, and one patient explicitly said she did not, and overall it was striking how many more patients reported blur as a key feature of their visual experience than a central blob. More broadly, patients felt that others didn't understand how faces looked to them, and also how variable this was (e.g., with lighting). Our Faces and Social Life in AMD Information Sheet provides a more accurate description of the range, and variability, of facial experience that patients might experience. This may be useful for medical staff explaining AMD to newly-diagnosed patients (e.g., orthoptists), for patients themselves, and also for family, friends and nurses to better understand why faces are so difficult for patients with AMD.

Finally, the reports of variability have important implications for the design of lab-based tests to assess objective face processing ability. Patients report that how well they see faces can vary substantially across lighting conditions, and across time (e.g., time of day, pre- vs. post- treatment with ranibizumab). This argues that, to obtain an accurate objective score for lab-based face ability, it may be important to test the patient on several different occasions, at different times of day, and with different lighting conditions on the faces.

Alternative strategies and importance of faces to successful real-world person recognition

The present study is the first in low vision to investigate the success or otherwise of non-face-based strategies for recognising others and their emotions. The key findings were that AMD patients report that, in everyday life, they attempt to use a wide range of alternative strategies (hairstyle, body shape, gait, voice, context) but these non-face-based strategies commonly fail. Specifically, we found that body language and tone of voice are not sufficiently strong cues to emotional state to fully enable normal social interactions. We also found that cues such as body shape, gait, and hairstyle are not sufficiently strong or reliable cues to identity to enable accurate recognition of who people are. This latter result is of some interest given occasional claims by vision science researchers that hairstyle is sufficient to support identity recognition (e.g., based on findings such as [62]). Our present results, in contrast, confirm previous findings from prosopagnosia (e.g., [17,35,36,63]), and also from low-resolution images (security camera video of walking people with faces covered [64]), that the ability to accurately perceive face information is crucial to reliable person recognition in everyday life, with other cues offering only partially useful information.

Generalisation to other low-vision conditions

Finally, our results are relevant to low-vision conditions beyond macular degeneration. As with AMD, studies discussing effects of other types of vision loss on social interactions, social life, and quality of life have not disentangled, in any detail, effects specifically related to face perception problems (e.g., [8,65,66]). The details of how faces appear to patients with different disorders will vary (e.g., see description of patient visual experience in glaucoma [67]). However, all vision disorders producing low visual acuity will result in problems seeing faces clearly. There is no reason to think that the types of difficulties in social interactions we have reported here are in any way related to the specific type of retinal damage that occurs in macular degeneration. Indeed, the (identity-related) social interaction difficulties in AMD closely mirror those present in prosopagnosia, where face recognition problems do not originate in the eye at all, but rather in the brain. Thus, it is highly likely that any eye disorder resulting in low vision will produce qualitatively similar social difficulties to those revealed in AMD, together with the same concomitant effects of missing out socially, tendency to social withdrawal, reduced confidence, and reduced quality of life specifically associated with face perception difficulties.