Early Performance of the Patients Over Paperwork Initiative among Family Medicine Physicians

Abstract

Objectives:

In 2019, the Centers for Medicare & Medicaid Services began implementing the Patients Over Paperwork (POP) initiative in response to clinicians’ reporting burdensome documentation regulations. To date, no study has evaluated how these policy changes have influenced documentation burden.

Methods:

Our data came from the electronic health records of an academic health system. Using quantile regression models, we assessed the association between the implementation of POP and clinical documentation word count using data from family medicine physicians in an academic health system from January 2017 to May 2021 inclusive. Studied quantiles included the 10th, 25th, 50th, 75th, and 90th quantiles. We controlled for patient-level (race/ethnicity, primary language, age, comorbidity burden), visit-level (primary payer, level of clinical decision making involved, whether a visit was done through telemedicine, whether a visit was for a new patient), and physician-level (sex) characteristics.

Results:

We found that the POP initiative was associated with lower word counts across all of the quantiles. In addition, we found lower word counts among notes for private payers and telemedicine visits. Conversely, higher word counts were observed in notes that were written by female physicians, notes for new patient visits, and notes involving patients with greater comorbidity burden.

Conclusions:

Our initial evaluation suggests that documentation burden, as measured by word count, has declined over time, particularly following implementation of the POP in 2019. Additional research is needed to see whether the same occurs when examining other medical specialties, clinician types, and longer evaluation periods.

Medical documentation is a critical component of clinical care, and it facilitates communication among the healthcare team.^1–5 For several decades, clinicians also have been tasked with recording additional information for billing purposes and to protect against medical malpractice claims.^1,2,6–9 Researchers have reported on the documentation burden (ie, time and effort clinicians spend on documentation)¹⁰ experienced among primary care physicians (PCPs) in the United States.^11–15 For instance, researchers estimate that for every hour of patient care that PCPs deliver, they spend an additional two hours on electronic health record (EHR) and administrative work.¹⁶ This likely contributes to the finding that PCPs are more apt to report burnout compared to specialists,^17–21 especially as growing evidence highlights an association between documentation burden and physician burnout.^{17,18,22–24} Unfortunately, unaddressed burnout may lead to turnover and lower quality of care.^25–28 Consequently, healthcare stakeholders are searching for solutions to address physician burnout and its causes, including documentation burden.^24,29–32

Starting in 2017, the Centers for Medicare & Medicaid Services (CMS) began developing a series of policy changes to documentation regulations through the Patients Over Paperwork (POP) initiative. These changes aimed to reduce clinicians’ documentation burden through several strategies, including reducing duplicative documentation requirements, streamlining documentation requirements, and revising federal payment rules.³³ In 2019, CMS began to implement the POP initiative. At the same time, they eliminated some duplicative documentation requirements, such as allowing clinicians to make statements that they reviewed information entered into the note by someone else (eg, medical students) as opposed to rewriting the information.³⁴ In 2020, CMS also simplified documentation requirements for certain services, such as standardizing documentation requirements for durable medical equipment.³⁵ In 2021, CMS revised federal payment rules for office visits to reduce the documentation burden associated with the Evaluation & Management (E&M) coding system, such as converting the five-level coding system into a three-level system.³⁴ The POP initiative may especially affect the documentation burden among certain physicians, such as PCPs. For example, prior researchers have highlighted that the E&M system does not align well with many forms of primary care services, such as chronic care management.³⁶ To our knowledge, however, no study has yet evaluated the association between the POP initiative and documentation burden among PCPs.

To address this research gap, our study examined how the POP initiative rollout was associated with documentation word count (a proxy of documentation burden) across family medicine physicians after controlling for other factors. The findings of this study may benefit policymakers who are optimizing the POP initiative.

Methods

We report this pooled cross-sectional study using the Strengthening the Reporting of Observational Studies in Epidemiology Statement.³⁷ The study period was January 2017 to May 2021 inclusive. The University of Florida’s institutional review board reviewed and approved the protocol.^A

Setting and Sample

Our study occurred at an academic health system in north-central Florida.^B We focused exclusively on visit notes written by family medicine physicians (ie, residents/fellows, attendings) because researchers have found that PCPs experience substantial documentation burden.^11–15 We excluded any note for a non-E&M service because we did not expect those visits to be affected by the POP initiative.³⁴ We used Current Procedural Terminology (CPT) codes (99201–99205 and 99211–99215) associated with each visit to determine whether a given visit was an E&M visit. We also excluded any note that was made in error, or stated that the patient had canceled or did not attend the visit. Because patient instructions often were made by templates with no changes by the physician, we also excluded these from analysis. We used data from the EHR system of the study site.

Measures

Dependent Variable

Our dependent variable was documentation length as measured by word count. For each unique visit made by a patient, we calculated the total word count across all of the notes associated with the visit. This included the chief complaint, Subjective, Objective, Assessment, and Plan sections of the note and teaching attestations made by attending physicians who supervised residents for the same visit.

Main Independent Variable

We were primarily interested in exposure to the POP initiative. Thus, we used the year of the visit date to determine when the visit occurred relative to when CMS implemented the POP initiative (ie, 2017–2018 preimplementation and 2019–2021 postimplementation periods).

Visit-Level Covariates

To control for other factors that may influence physicians’ documentation patterns, we included several variables that prior research had shown to be associated with documentation. This included whether the visit was conducted in person versus telemedicine and whether a medical student helped write the note.^38,39 To determine medical student involvement in note writing, we phrase-searched notes for “medical student note,” “MEDICAL STUDENT NOTE,” and “Medical Student Note” because we found that these phrases correctly flagged visits when a medical student was reportedly involved in note writing while correctly omitting visits in which the patient was a medical student or the medical student was only reported as shadowing (ie, did not assist in writing the note). We also assessed payer type to account for potential differences in documentation requirements and other systemic differences that may exist across payers. For instance, healthcare organizations may experience more billing issues (eg, repealing claim denials) with Medicaid than with other payer types.^40,41 We grouped patients who had Medicare, Medicaid, or TRICARE as having a public payer. We also treated visits insured by a Medicare Advantage plan as having a public payer. All of the visits insured by a private payer were treated as having a private payer. Lastly, we classified all other payers (eg, worker’s compensation) as “other.” No patients in this sample were uninsured or self-pay. We also hypothesized that notes for new visits may be longer than for return visits because of the need to document specific elements for the first time, such as personal and family medical history. Furthermore, longer visits and visits involving more intensive clinical decision making may lead to longer notes because of additional information the physician is considering and recording. Consequently, we included a dichotomous variable representing whether the visit was for a new patient or a return patient as well as a variable (ie, high level of service provided) representing whether the visit had a level 4 (99204, 99214) or level 5 (99205, 99215) CPT code.

Physician-Level Covariates

Because previous researchers have suggested differences in documentation patterns occur by physician sex,^42–45 we assessed for this.

Patient-Level Covariates

Other researchers have speculated that notes may be influenced by patient complexity, so we included age and comorbidity burden to account for this.^13,20 To measure comorbidity burden, we used the van Walraven scoring system because this system includes more comorbidities than the Charlson Comorbidity Index and adjusts for disease severity, unlike the Elixhauser system.⁴⁶ We used visit diagnoses and problem list entries to determine whether a given patient had a diagnosis for any of the disease concepts covered in the van Walraven system (eg, cardiac arrhythmias, depression).^46,47 The van Walraven adjusted scores can range from −16 to 92.⁴⁶ Further details on diagnosis codes and weights are described elsewhere.⁴⁶ Lastly, we hypothesized that other patient-level variables may influence documentation length. For instance, language barriers may limit the extent of medical information (eg, personal medical history) conveyed by the patient.⁴⁸ Thus, we also controlled for patient race/ethnicity and primary language.

Analytical Approach

We descriptively reported the sample characteristics and results of all Mann-Whitney U tests and Pearson χ² tests. Because we were interested in seeing how the independent variable and each of the covariates were associated across the distribution of the outcome variable, we conducted multivariable quantile regression models using the 10th, 25th, 50th, 75th, and 90th quantiles.^49,50 We used these approaches because the distribution of the outcome variable was skewed and we wanted to assess whether the POP initiative was associated with changes in word count across both shorter and longer notes. These five quantiles were selected based on previously published health services research studies that have used them.^51–54 Notably, the POP initiative has gradually implemented changes to documentation regulations since 2019.³⁴ As such, we created several dichotomous variables to represent each of the 5 years in our study period in our final models. We believe that this approach provides more detail than a dichotomous exposure (ie, pre- vs post-) variable and can allow us to evaluate whether a certain year was significantly associated with greater changes in documentation word count relative to other years. We assessed for multicollinearity and found that age at encounter was highly correlated with payer type. Consequently, our final models excluded age at encounter. Furthermore, we did not have enough variation in the variable representing whether a medical student was involved in note writing (ie, <99.9% did not involve a medical student). As such, our final models excluded this variable. Similar to other healthcare organizations, our study site began implementing telemedicine services in response to the coronavirus disease 2019 (COVID-19) pandemic.^55,56 We therefore ran an additional model that was restricted to notes in 2020 and 2021 and included an additional dichotomous variable representing whether a visit involved telemedicine to determine whether it was independently associated with documentation length. We used complete case analysis to address missing data. A P < 0.05 was interpreted as significant. All of the analyses were conducted with STATA SE 16.1 (StataCorp, College Station, TX) using the ‘qreg’ command to develop the quantile regression models, ‘strmatch’ syntax to phrase-search notes, and the ‘egen wordcount’ command to calculate the word count for each note.

Results

Our study included 32,790 visits. A median of 711 words were used across notes. Most of the visits were for return patients (89.5%), had a high level of clinical decision making involved (51.2%; ie, level 4 or 5 CPT code), were face-to-face visits (97.3%), and were insured by a public payer (57.7%). Most of the physicians studied were men (64.4%; Table 1).

Table 1.

Sample characteristics^A

Characteristic, n	Total sample size (N = 32,790)	Pre-POP (n = 6817)	Post-POP (n = 25,973)	P
Documentation word count, median (range)	711 (65–8683)	807 (65–7355)	688 (92–8683)	<0.001
Patient age at encounter, y, median (range)	49 (18–89)	52 (18–89)	49 (18–89)	<0.001
Patient race/ethnicity (%)				<0.001
Non-Hispanic White	15,865 (48.4)	3276 (48.1)	12,589 (48.5)
Non-Hispanic Black	13,828 (42.2)	3043 (44.6)	10,785 (41.5)
Hispanic	1709 (5.2)	267 (3.9)	1442 (5.6)
Non-Hispanic Asian	590 (1.8)	96 (1.4)	494 (1.9)
Non-Hispanic otherc	798 (2.4)	135 (2.0)	663 (2.6)
Primary language (%)				0.023
English	32,270 (98.4)	6688 (98.1)	25,582 (98.5)
Non-English	520 (1.6)	129 (1.9)	391 (1.5)
van Walraven comorbidity index, median (range)	0 (−13 to 32)	0 (−11 to 26)	0 (−13 to 32)	<0.001
Visit type (%)				0.061
Return	29,350 (89.5)	6144 (90.1)	23,206 (89.4)
New	3440 (10.5)	673 (9.9)	2767 (10.7)
High level of service provided (%)				<0.001
No	15,997 (48.8)	4134 (60.6)	11,863 (45.7)
Yes	16,793 (51.2)	2683 (39.4)	14,110 (54.3)
Telemedicine (%)				<0.001
No	31,889 (97.3)	6817 (100.0)	25,072 (96.5)
Yes	901 (2.8)	0 (0.0)	901 (3.5)
Payer (%)				<0.001
Public	18,930 (57.7)	4202 (61.6)	14,728 (56.7)
Private	13,652 (41.6)	2580 (37.9)	11,072 (42.6)
Other	208 (0.6)	35 (0.5)	173 (0.7)
Medical student helped write note (%)				0.105
No	32,780 (>99.9)	6817 (100.0)	25,963 (>99.9)
Yes	10 (<1)	0 (0.0)	10 (<1)
Physician sex (%)				<0.001
Male	21,116 (64.4)	5544 (81.3)	15,572 (60.0)
Female	11,674 (35.6)	1273 (18.7)	10,401 (40.1)
Visit year (%)				—
2017	2345 (7.2)	2345 (34.4)	—
2018	4472 (13.6)	4472 (65.6)	—
2019	8722 (26.6)	—	8722 (33.6)
2020	11,805 (36.0)	—	11,805 (45.5)
2021	5446 (16.6)	—	5446 (21.0)

^APlease indicate the locations in the Table 1 body of footnotes b and d (the footnotes also must be re-labeled alphabetically, beginning with a, and their order in the table body must be alphabetical).

POP, Patients Over Paperwork.

^bPercentages may not sum to 100% because of rounding.

^cNon-Hispanic other includes patients self-reporting as Native Americans, multiracial, or other.

^dPre-POP consisted of visits in calendar years 2017–2018 and post-POP consisted of visits in calendar years 2019 to mid-2021.

Detailed results of all of the variables in our adjusted model are displayed in Table 2. After controlling for other factors, we found that the years 2018, 2019, 2020, and 2021 were associated with lower word counts compared with 2017 across all of the quantiles. Notably, 2021 was consistently associated with the greatest reduction in word count across all of the quantiles. For instance, at the 50th quantile, 2018 was associated with 254.0 fewer words, 2019 was associated with 260.8 fewer words, and 2020 was associated with 322.5 fewer words when all were compared to 2017. Meanwhile, 2021 was associated with 415.6 fewer words compared with 2017.

Table 2.

Quantile regression parameter estimates for POP rollout and documentation word count from 2017 to mid-2021 (N = 32,790 visit notes)^B

Characteristic	10th quantile β (95% CI)	25th quantile β (95% CI)	50th quantile β (95% CI)	75th quantile β (95% CI)	90th quantile β (95% CI)
Year
2017	Ref	Ref	Ref	Ref	Ref
2018	−105.0 (−117.2 to −92.8)***	−150.3 (−161.7 to −138.9)***	−254.0 (−270.0 to −238.0)***	−294.9 (−322.7 to −267.1)***	−246.0 (−295.4 to −196.6)***
2019	−108.6 (−119.8 to −97.4)***	−161.5 (−172.0 to −151.0)***	−260.8 (−275.6 to −246.0)***	−280.1 (−305.7 to −254.5)***	−198.0 (−243.6 to −152.4)***
2020	−134.0 (−145.0 to −123.0)***	−199.5 (−209.8 to −189.2)***	−322.5 (−337.0 to −308.0)***	−352.7 (−377.8 to −327.6)***	−212.0 (−256.7 to −167.3)***
2021	−195.6 (−207.6 to −183.6)***	−268.8 (−280.0 to −257.5)***	−415.6 (−431.4 to −399.8)***	−512.3 (−539.7 to −484.9)***	−505.0 (−553.8 to −456.2)***
Patient race/ethnicity
Non-Hispanic White	Ref	Ref	Ref	Ref	Ref
Non-Hispanic Black	−2.0 (−7.5 to 3.5)	−1.9 (−7.0 to 3.3)	1.4 (−5.9 to 8.7)	−3.4 (−16.1 to 9.2)	19.0 (−3.5 to 41.5)
Hispanic	7.8 (−4.5 to 20.1)	8.0 (−3.5 to 19.5)	−1.1 (−17.3 to 15.1)	−22.8 (−50.9 to 5.3)	−68.0 (−118.0 to −18.0)**
Non-Hispanic Asian	5.8 (−14.6 to 26.2)	4.9 (−14.2 to 24.1)	−15.8 (−42.7 to 11.1)	−51.4 (−98.1 to −4.7)*	−98.0 (−181.1 to −14.9)*
Non-Hispanic otherb	2.0 (−15.5 to 19.5)	−2.3 (−18.7 to 14.1)	−10.2 (−33.2 to 12.8)	−17.0 (−59.9 to 22.9)	50.0 (−21.1 to 121.1)
Primary language
English	Ref	Ref	Ref	Ref	Ref
Non-English	−22.0 (−44.1 to 0.1)	−22.1 (−42.8 to −1.3)*	−2.6 (−31.8 to 26.6)	2.8 (−47.8 to 53.4)	−17.0 (−107.0 to 73.0)
van Walraven comorbidity index	5.6 (4.7-6.5)***	7.2 (6.4-8.1)***	10.9 (9.8-12.1)***	15.2 (13.1-17.2)***	19.0 (15.4-22.6)***
Visit type
Return	Ref	Ref	Ref	Ref	Ref
New	141.8 (133.2–150.4)***	154.7 (146.6–162.8)***	173.4 (162.0–184.8)***	190.6 (170.8–210.4)***	230.0 (194.8–265.2)***
High level of service provided
No	Ref	Ref	Ref	Ref	Ref
Yes	91.0 (85.7–96.3)***	97.1 (92.1–102.2)***	119.7 (112.7–126.7)***	162.6 (150.4–174.8)***	238.0 (216.3–259.7)***
Payer
Public	Ref	Ref	Ref	Ref	Ref
Private	−14.8 (−20.3 to −9.3)***	−24.6 (−29.7 to −19.4)***	−45.8 (−53.0 to −38.6)***	−98.0 (−110.5 to −85.5)***	−149.0 (−171.2 to −126.8)***
Other	−0.4 (−33.4 to 32.6)	23.5 (−7.4 to 54.4)	34.2 (−9.3 to 77.7)	10.5 (−65.0 to 85.9)	40.0 (−94.2 to 174.2)
Physician sex
Male	Ref	Ref	Ref	Ref	Ref
Female	3.4 (−2.2 to 9.0)	20.3 (15.1–25.5)***	33.3 (26.0–40.6)***	18.0 (5.2–30.7)**	−16.0 (−38.7 to 6.7)

^BPlease indicate the locations in the Table 2 body of footnotes c and d (the footnotes also must be re-labeled alphabetically, beginning with a, and their order in the table body must be alphabetical).

CI, confidence interval; POP, Patients Over Paperwork; Ref, reference.

^bNon-Hispanic other includes patients self-reporting as Native Americans, multiracial, or other.

^cAge at encounter was removed from the model because of multicollinearity with payer type.

^dR² = 0.0815 (10th quantile), 0.0851 (25th quantile), 0.0894 (50th quantile), 0.0892 (75th quantile), and 0.0807 (90th quantile).

^*P < 0.05.

^**P < 0.01.

^***P < 0.001.

Other notable covariates were associated with documentation word count in our model. Higher van Walraven comorbidity scores (ie, greater comorbidity burden), new visits, and high level of service billed were associated with higher word counts across all quantiles. Conversely, visits insured by a private payer were associated with lower word counts across all of the quantiles (Table 2).

When we restricted the model to only 2020 and 2021 data (Table 3), we observed similar results across the year effect and covariates as described in the previous paragraph. Across all of the quantiles, telemedicine was independently associated with lower word counts.

Table 3.

Quantile regression parameter estimates for POP rollout and documentation word count from 2020 to mid-2021 (N = 17,251 visit notes)^C

Characteristic	10th quantile β (95% CI)	25th quantile β (95% CI)	50th quantile β (95% CI)	75th quantile β (95% CI)	90th quantile β (95% CI)
Year
2020	Ref	Ref	Ref	Ref	Ref
2021	−54.7 (−60.9 to −48.4)***	−67.0 (−73.2 to −60.8)***	−87.1 (−95.7 to −78.5)***	−157.0 (−176.6 to −137.4)***	−297.5 (−322.7 to −262.3)***
Patient race/ethnicity
Non-Hispanic White	Ref	Ref	Ref	Ref	Ref
Non-Hispanic Black	−3.7 (−9.8 to 2.5)	−4.0 (−10.1 to 2.1)	−4.91 (−13.4 to 3.6)	−15.0 (−34.3 to 4.3)	−3.0 (−37.8 to 31.8)
Hispanic	2.3 (−10.6 to 15.3)	0.0 (−12.8 to12.8)	2.1 (−15.7 to 19.9)	−9.0 (−49.6 to 31.6)	−53.5 (−126.5 to 19.5)
Non-Hispanic Asian	−7.0 (−28.6 to 14.6)	−7.0 (−28.3 to 14.3)	−13.0 (−42.7 to 16.7)	−67.0 (−134.8 to 0.8)	−104.5 (−226.4 to 17.4)
Non-Hispanic otherb	8.7 (−9.8 to 27.1)	−5.0 (−23.2 to 13.2)	−21.8 (−47.2 to 3.5)	−12.25 (−70.1 to 45.6)	34.0 (−70.1 to 138.1)
Primary language
English	Ref	Ref	Ref	Ref	Ref
Non-English	3.7 (−22.5 to 29.8)	−3.0 (−28.8 to 22.8)	14.9 (−21.1 to 50.9)	20.0 (−62.1 to 102.1)	134.0 (−13.7 to 281.7)
van Walraven comorbidity index	4.6 (3.6–5.5)***	5.1 (4.2–6.1)***	6.6 (5.3–8.0)***	10.8 (7.7–13.8)***	16.3 (10.8–21.8)***
Visit type
Return	Ref	Ref	Ref	Ref	Ref
New	128.7 (118.9–138.4)***	158.0 (148.4–167.6)***	194.2 (180.8–207.6)***	204.8 (174.2–235.3)***	237.8 (182.8–292.8)***
High level of service provided
No	Ref	Ref	Ref	Ref	Ref
Yes	85.7 (79.7–91.6)***	99.0 (93.1–104.9)***	124.0 (115.8–132.2)***	166.0 (147.3–184.7)***	231.0 (197.4–264.6)***
Telemedicine
No	Ref	Ref	Ref	Ref	Ref
Yes	−86.7 (−99.8 to −73.6)***	−93.0 (−105.9 to −80.1)***	−104.2 (−122.2 to 86.2)***	−132.0 (−173.1 to −90.9)***	−177.0 (−251.0 to −103.0)***
Payer
Public	Ref	Ref	Ref	Ref	Ref
Private	−9.3 (−15.3 to −3.3)**	−18.0 (−23.9 to −12.1)***	−29.1 (−37.3 to −20.8)***	−66.0 (−84.8 to −47.2)***	−128.5 (−162.4 to −96.6)***
Other	10.3 (−23.9 to 44.6)	20.0 (−13.8 to 53.8)	24.0 (−23.2 to 71.2)	−4.3 (−111.8 to 103.3)	−11.0 (−204.5 to 182.5)
Physician sex
Male	Ref	Ref	Ref	Ref	Ref
Female	39.0 (33.1–44.9)***	49.0 (43.2–54.8)***	60.9 (52.8–69.0)***	45.3 (26.9–63.6)***	−7.5 (−40.6 to 25.6)

^CPlease indicate the locations in the Table 3 body of footnotes c and d (the footnotes also must be re-labeled alphabetically, beginning with a, and their order in the table body must be alphabetical).

CI, confidence interval; POP, Patients Over Paperwork; Ref, reference.

^bNon-Hispanic other includes patients self-reporting as Native Americans, multiracial, or other.

^cAge at encounter was removed from the model because of multicollinearity with payer type.

^dR² = 0.0935 (10th quantile), 0.0910 (25th quantile), 0.0859 (50th quantile), 0.0704 (75th quantile), and 0.0792 (90th quantile).

^*P < 0.05.

^**P < 0.01.

^***P < 0.001.

Discussion

This study assessed the association between the rollout of the POP initiative and documentation burden among family medicine physicians over time. Overall, we found that documentation word count did change during the POP rollout years 2019–2021. Specifically, when controlling for other factors, each year of the POP rollout period was associated with lower word counts compared with 2017. The notable covariates associated with higher word counts included greater comorbidity burden, new visits, higher clinical decision making involved, and female physicians. Conversely, covariates associated with lower word counts included being insured by a private payer and telemedicine visits. We highlight the implications below.

Our results suggest that the CMS POP initiative may have helped to partially mitigate documentation length. We observed a general pattern of a stepwise decrease in the number of words in the notes each year compared to the prior year. This encouraging finding suggests that the subsequent POP changes may be successfully building on the previous improvements from earlier changes. Interestingly, 2018 also was associated with significantly lower word counts despite the implementation of POP in 2019. This finding may suggest that other concurrent interventions, such as healthcare organizations’ EHR training and optimization efforts,⁵⁷ may be important contributors to reducing documentation burden. We also saw relatively shorter documentation lengths in 2021. This may provide support to earlier researchers’ speculations that changes in coding systems may need to be implemented before clinicians may report a noticeably lower documentation burden.⁵⁸ Our study included only 5 months of 2021 data, however, and serves as an initial evaluation of the set of changes that CMS implemented in January 2021. Subsequent research is needed to see whether the same findings on reduced word counts can be replicated with a longer evaluation period (eg, 1-year postimplementation). Of note, our study examined only one component of documentation burden, which may provide only one perspective on the immense documentation burden that PCPs report. Future research should consider other measures of documentation burden that were beyond the scope of the present study, such as after-hours documentation and timely documentation.¹⁰

Another key finding was that the word counts of medical documentation for privately insured patients were significantly lower compared with publicly insured patients. This finding suggests that there may be a ceiling effect observed when implementing interventions to reduce documentation length for notes for publicly insured patients. Systemic differences in documentation requirements have been noted across payers,⁴⁰ which may limit the extent of the benefit that PCPs can realize from the POP initiative. Future research should evaluate whether payer type influences clinicians’ documentation practices to gauge whether additional policy changes through POP would help to close the gap in documentation length among payers.

Our models also suggest that notes by female physicians may be longer than notes by male physicians. This finding adds to the growing body of literature that has found that female clinicians write longer notes and spend a greater amount of time managing inbox messages and writing notes.^42–45 Interestingly, our models revealed nuances where there were no significant differences by physician sex for notes that were shorter than the 10th quantile and longer than the 90th quantile. Additional research is needed to replicate these findings using larger datasets that encompass multiple healthcare organizations.

Lastly, we found that telemedicine visits were associated with shorter documentation length. This finding was intuitive as telemedicine visits are conducted differently compared with face-to-face visits. For instance, the Objective section of the note may be more brief because of the limited ability to complete a physical examination.^38,56,59 Because the ongoing COVID-19 pandemic coexisted with the POP initiative, however, it is unclear what proportion of the shortened note may be attributed to telemedicine visits versus POP. Further study may be needed to clarify these findings using healthcare organizations that were vanguards in telemedicine adoption before the onset of the ongoing COVID-19 pandemic.

Our study has several limitations. First, data came from one organization, limiting generalizability to other organizations. Second, we were unable to control for practice-level differences because some physicians practiced in additional locations. Third, some unmeasured variables, such as the extent of the note written by a template or copy-and-paste functions,^60,61 may influence documentation length. Future research should incorporate these measures when assessing the impact of interventions on documentation length. Fourth, our study was unable to account for patients without insurance or those who were self-pay, where notes may be less influenced by the documentation requirements of payers. Furthermore, our study focused on family medicine physicians. Differences in documentation length may exist by medical specialty. For instance, a study in ophthalmology found that documentation length had increased from 2006 to 2016.⁶¹ Future research should examine how POP affected documentation burden among other clinician types (eg, nurse practitioners) and other specialties that also perform E&M services (eg, general internal medicine, neurology). Lastly, we used an observational study design, precluding conclusive remarks on the causal impact of the POP initiative. Other concurrent organization-level interventions, such as EHR use optimization,⁶² may influence documentation length.

Conclusions

The POP was designed to reduce clinician documentation burden and ensure that clinicians can focus more on direct patient care. Our initial evaluation suggests that documentation burden, as measured by word count, has declined over time, particularly following implementation of the POP in 2019. Additional research is needed to see whether the same finding occurs when examining other medical specialties, clinician types, and longer evaluation periods.

Fig.

Timeline of Patients Over Paperwork. E&M, Evaluation & Management.^C

^CThe figure is not cited in the article. Please do so.

Key Points.

• When controlling for other factors, the word count for notes declined over time following the rollout of the Patients Over Paperwork initiative in 2019.

• The word count fell by approximately 25.7% in 2019 as compared with 2017.

• Other variables that were associated with higher documentation word count included greater comorbidity burden, new visits, higher clinical decision making involved, and female physicians.