Study design

This was a secondary analysis of a systematic random sampling study of KCMC EMD patients surveyed between October 2021 and May 2022. The primary study sought to describe gender differences in unhealthy alcohol use behaviors. As such, the EMD was chosen as a study location because of the high association between injuries and alcohol consumption [27, 39]. This current analysis seeks to compare alcohol consumption among IPs and NIPs who attend KCMC’s EMD. Thus, demographic factors, self-reported alcohol use, and AUDIT scores were assessed between these two patient populations. The SBIRT program matrix stands as this analysis’s guiding conceptual framework, given its aim of improving intervention implementation and effectiveness [40].

Study sample

Patients were enrolled at the KCMC EMD study site, which is situated in Moshi, Tanzania. All enrolled participants met the following eligibility criteria: 1) were 18 years of age or older, 2) had the capacity to give informed consent, 3) received initial care at KCMC Emergency Department, and 4) were conversant in Kiswahili. Patients who were prisoners were excluded from our study due to increased medicolegal risk, concern for power dynamics given their incarcerated state, and their limited access to alcohol during incarceration. For the safety of the data collection team, patients who tested positive for COVID-19 were also not approached. A patient was determined capable of providing informed consent if they were medically stabilized, clinically sober, and well enough to complete the survey verbally on their own. Those who were extremely ill or injured upon initial presentation were re-evaluated by the research team within 24 hours of arriving at KCMC or before discharge, whichever came first. Those who remained unable to consent within this time frame were excluded from study participation.

In addition to assessing gender differences, the original study aimed to compare rates of unhealthy alcohol use among females in the EMD and the reproductive health center, which required a higher proportion of female than male EMD subjects because of the lower hypothesized prevalence of alcohol misuse. Thus, the data analyzed here is representative of the EMD population when separated by gender but not when taken en masse. However, for this secondary analysis, sufficient subjects were enrolled overall to estimate the prevalence of AUDIT scores ≥ 8 (described more in-depth below) among IPs and NIPs within a 10% margin of error and 90% confidence level. Ethical approval for this study was obtained from the Kilimanjaro Christian Medical University College Ethics Review Board, the Tanzanian National Institute of Medical Research, and the Duke University Institutional Review Board prior to the initiation of data collection.

Data collection

Patients seeking care at KCMC’s EMD were enrolled Monday through Friday at peak attendance hours, 10:00 am until 6:00 pm local time. Patients who came in overnight, were initially intoxicated, or in critical condition were followed up over 24 hours to determine if they were able to provide informed consent. If so, and if agreeing to study participation, these patients were likewise enrolled. During the data collection period, enrollment of female patients remained consistent, but enrollment of male patients was paused from January 1st, 2021, until March 31st, 2022, awaiting regulatory approval of increased male sample size. To maintain a representative, systematic random sample and meet planned enrollment goals, every female but every third male presenting to the EMD for care was approached about study participation since the EMD sees significantly more males than females.

Patients who met eligibility criteria were screened and approached once medically stabilized and offered study participation. Research assistants, who were IRB approved and trained in Good Clinical Practices, explained the research protocol and informed consent to all interested patients. Written consent was formally denoted via a signature on an informed consent form for all interested patients, except for those who were illiterate, where, depending upon their ability level, consent was provided via initials or a cross-mark. Once enrolled, patients were asked questions related to their demographics, current alcohol intake, alcohol-related behaviors, consequences of their drinking, and depressive symptoms. All data were collected in Kiswahili by a team of three Tanzanian research assistants (two female and one male). For data collection, the male research assistant surveyed all male participants, and the female research assistants surveyed all female participants. This gender-matching was done to encourage open and honest reporting of patients’ experiences with alcohol based on local culture and research team experience [41]. Regulatory approval for the primary data collection project was obtained from the Kilimanjaro Christian Medical Center Ethics Committee, the Tanzanian National Institution of Medical Research (NIMR), and the Duke University Institutional Review Board. As much as possible, data was maintained in a de-identified manner and shared by data share agreement; personal health information was used for screening and enrollment, but data were collected, stored, and analyzed in a de-identified manner.

Variables

Information on patients’ demographics (sex, age, years of education, type of employment, income, marital status, tribe, and religion), injury status, and self-reported alcohol use was collected. All demographic and self-reported alcohol use questions had previously been translated into Kiswahili by our Tanzanian research team. The Alcohol Use Disorder Identification Test (AUDIT), Drinkers’ Inventory of Consequences, and the Patient Health Questionnaire 9 (PHQ-9) were the three relevant scales administered.

AUDIT is a commonly used 10-question tool for measuring alcohol consumption and alcohol-related problems, incorporating metrics that assess the quantity, frequency, and consequences of alcohol use [42, 43]. This multi-layered approach to quantifying the content of alcohol consumed makes AUDIT more sensitive and accurate in a multicultural setting, which is why it was selected as a primary measure in this study. AUDIT scores range from 0 to 40, with lower values indicating low-risk consumption and higher values suggesting alcohol dependence. Both locally and globally, scores greater than or equal to 8 are clinically significant for harmful or hazardous drinking (HHD), meaning an individual is at risk for suffering adverse health outcomes as a result of their alcohol intake [43–48]. AUDIT scores greater than or equal to 8 were thus used as a cut-off point in this analysis while also triggering the initiation of further clinical treatment for alcohol misuse.

DrInC (which ranges from 0–50) is a 50-question survey measuring alcohol-related consequences across five domains: interpersonal, intrapersonal, social consequences, impulse control, and physical [49]. While no clinically significant cut-off score exists, higher scores indicate more significant consequences for an individual [50]. Finally, PHQ-9 is a diagnostic tool used to identify the existence and severity of depression [51]. This scale ranges from 0 to 27, with higher values indicating increasingly severe depressive symptoms. Scores of 9 or greater were found to be the optimal cut-off score for identifying clinical depression in the KiSwahili-translated version of the PHQ-9 [52] and thus was the cut-off point used here. All three scales had previously been cross-culturally adapted, psychometrically validated, and clinically tested in the local context [48, 52, 53].

Data analysis

Data on patients’ demographics, injury status, self-reported alcohol use, and AUDIT, DrInC, and PHQ-9 scores were analyzed using descriptive frequencies and proportions. All variables were categorical except for measures of income, education status, and the three survey tools, which were continuous and were analyzed as means with standard deviation. AUDIT and PHQ-9 scores were dichotomized according to the cut-off values discussed previously; AUDIT scores of 8 and greater were categorized as HHD, while scores less than 8 were classified as not HHD. PHQ-9 scores of 9 and greater were categorized as a positive screen for depression, and scores less than 9 were classified as screening negative. Wilcoxon rank sum testing, Pearson’s Chi-squared testing, and Fisher’s exact testing were performed to determine statistical significance as appropriate. Missing data was minimal at 1 to 2 missing data points per variable, except for personal and household income as several participants were hesitant to disclose their financial status to the research staff. All data were analyzed in RStudio (version 1.4) using user-created and validated R-Packages.

Patient demographics

During the eight months of data collection from October 1st until May 31st, 376 EMD patients were surveyed, of which 59 (15.78%) presented with injuries and 315 (84.22%) presented for non-injury-related reasons, including but not limited to fever, headache, stomachache, body numbness, or body swelling. As expected, based on our sampling strategy, females account for 70.32% of the data collected. Across all groups, most patients were Christian and from the Chagga tribe (77.54% and 46.93%, respectively). Roughly half (46.13%) of participants were employed at the time of enrollment (10.40% were students and held no other employment, and 43.20% were unemployed). This demographic data can be seen in Table 1 below.

Download:

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

Table 1. Demographics of EMD patients by injury status.

https://doi.org/10.1371/journal.pgph.0001900.t001

Alcohol use characteristics of EMD patients

In following the intention of this paper to compare rates of alcohol use and alcohol-related behaviors across IPs and NIPs presenting to KCMC’s EMD, differences in AUDIT, PHQ-9, and DrInC scores were examined across these two groups. IPs had higher average [SD] AUDIT scores (5.76 [6.6]) than NIPs (3.93 [6.1]), although a sizeable proportion of individuals with HHD (AUDIT ≥ 8) were present in both patient groups (37.29% of IPs and 20.63% of NIPs, p<0.01). Average DrInC scores were higher in the IP population (12.92 [17.54]) than the NIP population (9.19 [15.99]), but mean PHQ-9 scores were greater among NIPs (6.60 [5.17]) than IPs (4.12 [4.14], p <0.01). Following this, a greater percentage of NIPs (21.90% of NIPs and 8.47% of IPs, p = 0.02) screened positively for depression (PHQ-9 ≥ 9).

Other markers of alcohol use, including drinking frequency and quantity, were also explored. AUDIT scores were higher among IPs; however, a greater proportion of NIPs appeared to drink more frequently (5.07% of NIPs but 1.7% of IPs reportedly drank every day or multiple times per day) and in unhealthy quantities (5.1% of NIPs but 3.4% of IPs drank five or more bottles per sitting). Other characteristics were comparable across both patient groups; roughly half (47.46% of IPs and 51.52% of NIPs, p = 0.93) had previously attempted to quit drinking, and a tenth (13.56% of IPs and 8.89% of NIPs, p = 0.27) had sought treatment for alcohol use, with personal reasons (39.29% of IPs and 41.72% of NIPs) cited as the primary motivation to quit. Most participants noted beer (30.51% of IPs and 23.81% of NIPs) as their alcoholic drink of choice, and 1.67% of IPs and 1.90% of NIPs spent more than 50,000 TZS (21.42 USD) on alcohol per week. To put this amount into context, the average reported personal income for the study population is 254,939.87 TZS (109.32 USD) per month or 58,674.31 TZS (25.16 USD) per week. These findings can be seen in Table 2.

Download:

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

Table 2. Alcohol use characteristics of EMD patients by injury status.

https://doi.org/10.1371/journal.pgph.0001900.t002

This data was separated further by males and females to account for how the sex-based differences in alcohol use could impact trends among IPs and NIPs. Of those with injuries, roughly half were females (44.07%), compared to the non-injury group, where females constituted the majority (75.24%) (Table 1). Overall, males had higher average AUDIT and DrInC scores but lower PHQ-9 scores than female patients, regardless of injury status. The average AUDIT score was 7.58 [7.27] for male IPs and 6.68 [8.59] for male NIPs. This stands in comparison to females; 3.46 [5.03] and 3.03 [4.76] were the average AUDIT score for female IPs and NIPs, respectively, with p<0.01. These trends are echoed in the proportion of patients who tested positive for HHD. For males, 48.48% of those with injuries and 34.62% of those without injuries screened positively for HHD, whereas for females, 23.08% of those with injuries and 16.03% of those without injuries had AUDIT scores greater than or equal to 8, with p<0.01.

Males also had average DrInC scores (16.30 [19.32] for male IPs and 14.28 [20.46] for male NIPs), which were higher than their female counterparts. For females, average DrInC scores were 8.62 [14.19] for female IPs and 7.49 [13.90] for female NIPs. While males had the highest AUDIT and DrInC scores, females were more likely to screen positively for depression (those with a PHQ-9 ≥ 9 constituted 15.38% of female IPs and 25.74% of female NIPs) compared to males (3.03% of male IPs and 10.26% of male NIPs had a PHQ-9 ≥ 9). Of all groups, female NIPs had the highest average PHQ-9 score (7.30 [5.16]), followed by female IPs with an average of 6.62 [3.94]. These differences were all statistically significant (p<0.01). These reports and others can be seen in Table 3.

Download:

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

Table 3. Alcohol use characteristics of EMD patients by injury status and gender.

https://doi.org/10.1371/journal.pgph.0001900.t003

Strengths and limitations

Several important factors should be considered when interpreting the data presented in this paper. The purposeful oversampling of female patients in the study from which this data originates impacted our reported incidence of alcohol use when looking at amalgamated IP and NIP statistics. This resulted in an underestimation of true IP and NIP alcohol intake in Table 2. To account for this bias, we have also reported statistics separated by gender, as seen in Table 3. Further, a complete registry of all patients who present for services at KCMC is unavailable, which restricts our ability to evaluate how representative this sample is of the EMD’s total population. While this limitation is offset by our large sample size and systematic random sampling strategy, it should still be taken into account. Our data set is also limited in terms of specificity around patients’ chief complaints. This constrains our ability to compare complaint-specific groups, which could help assess the risk of alcohol misuse in particular EMD patient groups. Therefore, we recommend that future studies look more closely at EMD patients’ demographics and chief complaints in combination with their alcohol intake, as it might vary with each unique setting and characteristic.

Some additional factors may have altered the composition of enrolled patients from what may ordinarily be seen of the full EMD population. First, our data was collected throughout several waves of the COVID-19 pandemic. The limited resources available to protect KCMC’s EMD patients from the spread of COVID-19 may have dissuaded those with lower acuity complaints from seeking care at the EMD, leading our sample to encompass a higher proportion of patients with severe conditions. Further, patients who remained unable to consent over 24 hours were excluded from study participation. While this was deemed necessary for ethical and safety reasons, it may have impacted study demographics given potentially different alcohol use patterns in this patient group than those who could provide consent. Last, the data collected from participants was dependent upon the accuracy of their self-reported demographic and alcohol intake. While several measures were taken throughout the study to encourage truthful reporting of patients’ typical alcohol intake (including gender-matching data collectors with participants and collecting all data in private, confidential environments), there may have been an underreporting of true alcohol consumption. This may be due to the societal stigmatization surrounding heavy alcohol use in Tanzania, which encourages secretive alcohol use behaviors, especially among females.

More research is needed to identify better treatments that would be effective for all EMD patients presenting with unhealthy alcohol use. Building upon the information presented here, to reduce the burden caused by alcohol in this region, more effective, evidence-based care options are urgently needed.