Abstract
This paper studies physician workflow management in primary care clinics using terminating Markov chain models. The physician workload is characterized by face-to-face encounters with patients and documentation of electronic health record (EHR) data. Three workflow management policies are considered: preemptive priority (stop ongoing documentation tasks if a new patient arrives); non-preemptive priority (finish ongoing documentation even if a new patient arrives); and batch documentation (start and finish documentation when the desired number of tasks is reached). Analytical formulas are derived to quantify the performance measures of three management policies, such as physician’s daily working time, patient’s waiting time, and documentation waiting time. A comparison of the results under three policies is carried out. Finally, a case study in a primary care clinic is carried out to illustrate model applicability. Such a work provides a quantitative tool for primary care physicians to design and manage their workflow to improve care quality.
Similar content being viewed by others
References
American College of Physicians (2006) The impending collapse of primary care medicine and its implications for the State of the Nation’s Health Care. Washington DC
Bodenheimer T (2006) Primary care—will it survive? New Engl J Med 355(9):861–864
Sinsky C A, et al. (2013) In search of joy in practice: a report of 23 high-functioning primary care practices. Ann Fam Med 11(3):272–278
Dyrbye L N, Shanafelt T D (2011) Physician burnout: a potential threat to successful health care reform. JAMA 305(19):2009– 2010
Sinsky C, et al. (2016) Allocation of physician time in ambulatory practice: a time and motion study in 4 specialties. Ann Intern Med 165(11):753–760
Arndt B G, et al. (2017) Tethered to the EHR: primary care physician workload assessment using EHR event log data and time-motion observations. Ann Fam Med 15(5):419–426
Young R, et al. (2018) A time-motion study of primary care physicians’ work in the electronic health record era. Fam Med 50(2):91–99
Pizziferri L, et al. (2005) Primary care physician time utilization before and after implementation of an electronic health record: a time-motion study. J Biomed Inform 38(3):176–188
Beasley J W, et al. (2011) Information chaos in primary care: implications for physician performance and patient safety. J Am Board Fam Med 24(6):745–751
Bae J, Encinosa W E (2016) National estimates of the impact of electronic health records on the workload of primary care physicians. BMC Health Serv Res 16(1):172
Wetterneck T B, et al. (2012) Development of a primary care physician task list to evaluate clinic visit workflow. BMJ Qual Saf 21(1):47–53
Holman G T, et al. (2016) The myth of standardized workflow in primary care. J Am Med Inf Assoc 23(1):29–37
Mishra P, et al. (2018) Association of medical scribes in primary care with physician workflow and patient experience. JAMA Intern Med 178(11):1467–1472
Brandeau ML et al (2004) Operations research and health care: a handbook of methods and applications. Springer, Berlin
Hall R W (2006) Patient flow: reducing delay in healthcare delivery. International series in operation research and management science. Berlin, Springer
Yang H, Lee EK (2016) Healthcare analytics: from data to knowledge to healthcare improvement. Wiley, New York
Yih Y (2016) Handbook of healthcare delivery systems. CRC Press, Boca Raton
Li J et al (2017) Stochastic models in healthcare delivery systems. World Scientific Publishing, Singapore
Jacobson SH et al (2006) Discrete-event simulation of health care systems. In: Hall RW (ed) Patient flow: reducing delay in healthcare delivery, vol 91, pp 211–252
Fomundam S, Herrmann J W (2007) A survey of queuing theory applications in healthcare. ISR Technical Report 2007-24 University of Maryland, College Park
Gupta D, Denton B (2008) Appointment scheduling in health care: challenges and opportunities. IIE Trans 40(9):800–819
Denton B T, et al. (2011) Medical decision making: open research challenges. IIE Trans Healthc Syst Eng 1(3):161–167
Zhong X et al (2016) Discrete event simulation for primary care redesign: review and a case study. In: Yang H, Lee EK (eds) Healthcare analytics: from data to knowledge to healthcare improvement, pp 399–426
Xie H, et al. (2005) A continuous time Markov model for the length of stay of elderly people in institutional long-term care. J R Stat Soc: Ser A (Stat Soc) 168(1):51–61
Cote M J, Stein W E (2007) A stochastic model for a visit to the doctor’s office. Math Comput Model 45(3–4):309–323
Wang J, et al. (2012) Modeling and analysis of work flow and staffing level in a computed tomography division of University of Wisconsin Medical Foundation. Health Care Manag Sci 15(2):108–120
Zhong X, et al. (2016) Analysis and design of gastroenterology (GI) clinic in digestive health center: a systems approach. Flex Serv Manuf 28(1–2):90–119
Zhong X, et al. (2016) A system theoretic approach to modeling and analysis of mammography testing process. IEEE Trans Syst Man Cybern: Syst 46(1):126–138
Zhong X, et al. (2018) Workload balancing—staffing ratio analysis for primary care redesign. Flex Serv Manuf J 30(1–2):6–29
Lee H K, et al. (2018) Joint visit in primary care clinics: modeling, analysis, and an application study. IISE Trans Healthc Syst Eng 8(2):93–109
Green L V, Savin S (2008) Reducing delays for medical appointments: a queueing approach. Oper Res 56(6):1526–1538
Jiang L, Giachetti R E (2008) A queueing network model to analyze the impact of parallelization of care on patient cycle time. Health Care Manag Sci 11(3):248–261
Liu N, D’Aunno T (2012) The productivity and cost-efficiency of models for involving nurse practitioners in primary care: a perspective from queueing analysis. Health Serv Res 47(2):594–613
Liu N, et al. (2014) A new model for nurse practitioner utilization in primary care: Increased efficiency and implications. Health Care Manag Rev 39(1):10–20
Zeng B, et al. (2013) The impact of overbooking on primary care patient no-show. IIE Trans Healthc Syst Eng 3(3):147–170
Zhong X, et al. (2017) Electronic visit in primary care: modeling, analysis, and scheduling policies. IEEE Trans Autom Sci Eng 14(3):1451–1466
Zhong X, et al. (2018) The impact of e-visit on patient access to primary care. Health Care Manag Sci 21(4):475–491
Chen N, et al. (2018) An analytical framework for modeling, analysis and improvement of team communication and collaboration process in primary care clinics. IEEE Trans Autom Sci Eng 16(3):1148–1162
Hahn J S, et al. (2012) Rapid implementation of inpatient electronic physician documentation at an academic hospital. Appl Clin Inform 3(2):175–185
Li J, Meerkov S M (2005) On the coefficients of variation of up- and downtime of manufacturing equipment. Math Probl Eng 2005:1–6
Li J, Meerkov SM (2009) Production systems engineering. Springer, Berlin
Nielsen BF (2017) Lecture notes on phase–type distributions for 02407 Stochastic Processes, http://www2.imm.dtu.dk/courses/02407/lectnotes/ftf.pdf, Technical University of Denmark, Lyngby, Denmark
Acknowledgments
The authors thank to the staff at the collaborating clinic for their support and help in this study. In addition, the study was conducted under the approval and management of human subject committee of the collaborating clinic and followed the HIPPA protocols and ethical guidelines. Since the analysis only involved workflow time stamps, no patient information was collected.
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This paper is supported in part by NSF Grant CMMI-1536987.
Appendices
Appendix A: Transitions
In PEP model, the transitions occur based on the following conditions:
-
(P1)
If \({X_{p}^{k}} = {X_{d}^{k}} = 0\) and \({X_{c}^{k}} <N\), the physician is idle (no patient and documentation are in process or waiting); thus only Transition 1 can occur.
-
(P2)
If \({X_{p}^{k}} > 0\) and \({X_{p}^{k}} + {X_{d}^{k}} +{X_{c}^{k}} < N\), there are patients in the system so that the physician is meeting with them; thus both Transitions 1 and 2 can occur. But Transition 3 cannot occur because the physician is working on a patient.
-
(P3)
If \({X_{p}^{k}} > 0\) and \({X_{p}^{k}} + {X_{d}^{k}} + {X_{c}^{k}} =N\), there are patients in the system and no more patient will come; thus only Transition 2 can occur.
-
(P4)
If \({X_{p}^{k}} = 0\), \({X_{d}^{k}} > 0\) and \({X_{d}^{k}} + {X_{c}^{k}}< N\), no patient is in the system, but documentation is not finished and still more patients will come; thus both Transitions 1 and 3 can occur.
-
(P5)
If \({X_{p}^{k}} = 0\), \({X_{d}^{k}} > 0\) and \({X_{d}^{k}} + {X_{c}^{k}} =N\), no patient is in the system, documentation is not done yet but no more patient will come; thus only Transition 3 can occur.
In NPP model, the transitions occur based on the following rules:
-
(N1)
If \({X_{p}^{k}} = {X_{d}^{k}} = 0\) and \({X_{c}^{k}} < N\), the physician is idle; thus only Transition 1 can occur.
-
(N2)
If \({D_{d}^{k}} = 0\), \({X_{p}^{k}} > 0\) and \({X_{p}^{k}} + {X_{d}^{k}} + {X_{c}^{k}} < N\), there are patients in the system, and more will come, but no documentation is in process; thus both Transitions 1 and 2 can occur.
-
(N3)
If \({D_{d}^{k}} = 0\), \({X_{p}^{k}} > 0\) and \({X_{p}^{k}} + {X_{d}^{k}} + {X_{c}^{k}} = N\), there are patients in the system, but no patient will come and no documentation is in process; thus only Transition 2 can occur.
-
(N4)
If \({D_{d}^{k}} = 0\), \({X_{p}^{k}} = 0\), \({X_{d}^{k}} > 0\) and \({X_{d}^{k}} + {X_{c}^{k}} < N\), no patient is in the system but more will come, and documentation is in process; thus both Transitions 3 and 4 can occur.
-
(N5)
If \({D_{d}^{k}} = 0\), \({X_{p}^{k}} = 0\), \({X_{d}^{k}} > 0\) and \({X_{d}^{k}} + {X_{c}^{k}}= N\), all patients have been served and documentation is in process; thus only Transition 3 can occur.
-
(N6)
If \({D_{d}^{k}} = 1\) and \({X_{p}^{k}} + {X_{d}^{k}} + {X_{c}^{k}} < N\), a new patient comes while documentation is in process, and more patients will come; thus both Transitions 3 and 4 can occur.
-
(N7)
If \({D_{d}^{k}} = 1\) and \({X_{p}^{k}} + {X_{d}^{k}} + {X_{c}^{k}} = N\), the last patient arrives while documentation is in process; thus only Transition 3 can occur.
In BDC model, the transitions occur in the following scenarios:
-
(B1)
If \(({X_{c}^{k}} \bmod M) = 0\), \({X_{p}^{k}} = 0\), \({X_{d}^{k}} < M\) and \({X_{p}^{k}} +{X_{d}^{k}} + {X_{c}^{k}} < N\), no patient is in the system, and previous documentation batch is finished but current one is not started since the batch is not filled, so the physician is idle; thus only Transition 1 can occur.
-
(B2)
If \(({X_{c}^{k}} \bmod M) = 0\), \({X_{p}^{k}} > 0\), \({X_{d}^{k}} < M\) and \({X_{p}^{k}} +{X_{d}^{k}} + {X_{c}^{k}} < N\), there are patients in the system, and previous documentation batch is finished but current one is not started due to less number of tasks in the batch, so the physician is meeting with a patient; thus both Transitions 1 and 2 can occur.
-
(B3)
If \(({X_{c}^{k}} \bmod M) = 0\), \({X_{d}^{k}} = M\) and \({X_{p}^{k}} + {X_{c}^{k}} <N-M\), the documentation batch size is reached so the physician will start working on the batch, and more patients will come; thus both Transitions 1 and 3 can occur.
-
(B4)
If \(({X_{c}^{k}} \bmod M) = 0\), \({X_{d}^{k}} = M\) and \({X_{p}^{k}} + {X_{c}^{k}} =N-M\), the documentation batch size is reached so the physician will start working on the batch, but no patient will come; thus only Transition 3 can occur.
-
(B5)
If \(({X_{c}^{k}} \bmod M) = 0\), \({X_{p}^{k}} > 0\), \({X_{d}^{k}} < M\) and \({X_{p}^{k}} +{X_{d}^{k}} + {X_{c}^{k}} = N\), there are patients in the system, and previous documentation batch is finished and current one is not started due to less number of tasks in the batch, so the physician is meeting with a patient, but no new patient will come; thus only Transition 2 can occur.
-
(B6)
If \(({X_{c}^{k}} \bmod M) \neq 0\) and \({X_{p}^{k}} + {X_{d}^{k}} + {X_{c}^{k}} < N\), the current documentation batch is in process and more patients will come; thus both Transitions 1 and 3 can occur.
-
(B7)
If \(({X_{c}^{k}} \bmod M) = 0\), \({X_{p}^{k}} = 0\) and \({X_{d}^{k}} + {X_{c}^{k}} = N\), all patients have been served, so the physician will work on documentation of the last batch no matter the batch size is not reached; thus only Transition 3 can occur.
-
(B8)
If \(({X_{c}^{k}} \bmod M) \neq 0\) and \({X_{p}^{k}} + {X_{d}^{k}} + {X_{c}^{k}} = N\), the current documentation batch is in process and no new patient will come; thus only Transition 3 can occur.
For the transition rate, ∀l = 0, 1,…,N − 1; n = 0, 1,…,N − l − 1; m = 0, 1,…,N − l − n − 1, in the PEP model, we have the following transition rates corresponding to Transitions 1 to 3 described in (P1)-(P5):
In the NPP model, it follows from Transitions 1 to 3 explained in (N1)–(N7) that
In the BDC model, from Transitions 1 to 3 outlined in (B1)–(B8), we obtain,
Appendix B: Proofs
Proof Proof of Lemma 1
For a given number of completed documentation task l, the number of patients in the system, m, ranges from 0 to N − l. With given l and m, the number of unfinished documents is from 0 to N − l − m (i.e., N − l − m + 1 cases). Thus,
For l ∈ [0,N], we obtain
□
Proof Proof of Lemma 2
The state space constraints for the number of patients in the system m, the number of unfinished documentation tasks n and the number of documentations tasks l in the NPP model are identical to those in the PEP model. Therefore, when \({D_{d}^{k}}=0\), the number of feasible states equals to KPEP. When Dd = 1, m and n cannot be 0 due to the constraints of Dd. The number of states for m = 0 or n = 0 is
Thus, we have
□
Proof of Lemma 3
For a given number of completed documentation tasks l, when l is a multiple of the batch size M, the number of unfinished documentation tasks n varies from 0 to \(\min \limits (M,N-l)\). Furthermore, with a fixed n, the number of patients in the system m varies from 0 to N − n − l (i.e., N − n − l + 1 cases).
Let \(A = \lfloor \frac {N}{M}\rfloor \) and B = N − AM. If l = Mk, where k is a non-negative integer, then
If l = Mk + j where j is a positive integer and j < M, then, n takes M − j so that n < M and ((n + l) mod M) = 0 due to state space constraints, and m varies from 0 to N − M(1 + k). Therefore, when \(Mk<l<\min \limits (M(1+k),N)\), the number of feasible states is (M − 1)(N + 1 − M(1 + k)) if k < A, and B if k = A. Thus, we can obtain
□
Appendix C: Algorithms for C V = 0
To evaluate the performance measures when CV = 0, calculation algorithms can be introduced for each model. In these algorithms, ta, ts, and td denote inter-arrival, service, documentation times, and N and M represent total number of patients and documentation batch size, respectively. Note that closed formulas for \(T_{0}^{\text {PEP}}\), \(T_{0}^{\text {NPP}}\), \(T_{0}^{\text {BDC}}\) and \(W_{0}^{\text {PEP}}\) can be obtained through the algorithm.
Rights and permissions
About this article
Cite this article
Lee, S., Bain, P.A., Musa, A.J. et al. A Markov chain model for analysis of physician workflow in primary care clinics. Health Care Manag Sci 24, 72–91 (2021). https://doi.org/10.1007/s10729-020-09517-1
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10729-020-09517-1