Multisite analysis of patient experience scores and risk of hospital admission: a retrospective cohort study

Abstract

Background:

Routinely collected patient experience scores may inform risk of patient outcomes. The objective of the study was to evaluate the risk of hospital admission within 30-days following third-party receipt of the survey and guide interventions.

Methods:

In this retrospective cohort study, we analyzed Hospital Consumer Assessment of Healthcare Providers and Systems surveys, January 2016–July 2019, from an institution’s 20 hospitals in four U.S. states. Surveys were routinely sent to patients using census sampling. We analyzed surveys received ≤60 days following discharge from patients living ≤60 miles of any of the institution’s hospitals. The exposures were 19 survey items. The outcome was hospital admission within 30-days after third-party receipt of the survey. We evaluated the association of favorable(top-box) vs unfavorable(non–top-box) score for survey items with risk of 30-day hospital admission in models including patient and hospitalization characteristics and reported adjusted odds ratios (aOR[95% confidence interval]).

Results:

Among 40,162 respondents (mean age±standard deviation: 68.1±14.0 years), 49.8% were women and 4.3% had 30-day hospital admission. Patients with 30-day hospital admission, compared to those not admitted, were more likely to be discharged from a medical service line (62.9% vs 42.3%;P<0.001) and have a higher Elixhauser index. Favorable vs unfavorable score for hospital rating was associated with lower odds of 30-day hospital admission in the overall cohort (0.88[0.77–0.99];P=0.04), medical service line (0.81[0.70–0.94];P=0.007), and upper tertile of Elixhauser index (0.79[0.67–0.92];P=0.003). Favorable score for recommend hospital was associated with lower odds of 30-day hospital admission in the medical service line (0.83[0.71–0.97];P=0.02) but for others (e.g., cleanliness of hospital environment) showed no association.

Conclusion:

In routinely collected patient experience scores, favorable hospital rating was associated with lower odds of 30-day hospital admission and may inform risk stratification and interventions. Evidence-based survey items linked to patient outcomes may also inform future surveys.

Introduction

Patient evaluation of hospital experience promotes transparency in patient care, quality improvement, and accountability in healthcare.[1] The Hospital Consumer Assessment of Healthcare Providers and Systems (HCAHPS) survey was developed by the Centers for Medicare & Medicaid Services (CMS) and the Agency for Healthcare Research and Quality and collects patient experience data following hospital discharge. The survey serves as a valuable resource of publicly available patient experience data in the United States.[2] Patients receive HCAHPS surveys 2–42 days following hospital discharge and may return them to a third-party vendor up to 90 days following discharge. Patient experience scores relate to the quality of care received during hospitalization and may also identify risk for outcomes following hospital discharge.

The sparse literature on patient experience scores and outcomes following hospital discharge has limited the wider use of routinely collected patient experience data. Previous studies on patient experience and outcomes focused on 30-day hospital readmission rates (i.e., readmission within 30 days of hospital discharge) and yielded mixed results.[3-9] In a single center study, hospitalized patients who were very satisfied with care had 39% lower odds of 30-day hospital readmission.[6] Patients with low trauma during hospitalization (based on disturbances in sleep, mobility, nutrition, and mood) had lower risk of 30-day hospital readmission.[9] However, in a national study, higher patient satisfaction was associated with 12% higher odds of hospital admission.[4] To our knowledge, there are no reports on HCAHPS surveys and patient outcomes >30 days following hospital discharge, thereby, underutilizing surveys returned >30 days of discharge and missing opportunities to evaluate additional patient outcomes.

Therefore, to leverage routinely conducted HCAPHS survey data, we conducted a retrospective study at an academic institution’s 20 hospitals across four U.S. states on the association of HCAHPS scores and a novel outcome, namely, hospital admission within 30 days of third-party vendor receipt of the survey. The objective of the study was to describe the respondents and evaluate the association of HCAHPS scores with 30-day hospital admission.

Methods

Study Design and Participants

Participants were patients who returned the HCAHPS survey following hospitalization at Mayo Clinic hospitals in Phoenix, Arizona; Jacksonville, Florida; and Rochester, Minnesota; and Mayo Clinic Health System (MCHS) hospitals in Minnesota and Wisconsin. The study sites (Arizona, Florida, Rochester, MCHS) were randomly labelled A–D. MCHS is a network that included 17 hospitals in Minnesota (Albert Lea/Austin, Cannon Falls, Fairmont, Lake City, Mankato, New Prague, Springfield, St. James, Red Wing, and Waseca) and Wisconsin (Barron, Bloomer, Eau Claire, La Crosse, Menomonie, Osseo, Sparta). Surveys were sent in English, Spanish, or Arabic, based on language preference in the electronic medical record. Using a retrospective cohort design, we analyzed surveys from patients discharged between January 1, 2016 and July 30, 2019.

We reported the study in accordance with American Association for Public Opinion Research (AAPOR) guidelines.[10]

HCAHPS Survey Instrument

Using census sampling, a third-party vendor (Press-Ganey) sent HCAHPS surveys to eligible patients 2–42 days following hospital discharge. Patients did not receive surveys if (i) aged <18 years at admission, (ii) hospitalized for ≤1 night, (iii) in legal custody, (iv) residents of a non-US country, (v) discharged to hospice care, nursing home, or skilled nursing facility, and/or (vi) included in a ‘no publicity’ and do not survey list. The third-party vendor contacted up to 100% of eligible patients. We received 107,822 surveys (25%–32% response rate per quarter; unpublished); data on patients sampled and characteristics of non-respondents were unavailable for analysis, as previously described.[11] The response rate was slightly higher than the national average of 24%.[12]

Sampling Process and Survey Item Processing

Of 107,822 surveys, we included surveys of patients discharged from medical or surgical service lines [13] and the first survey if >1 survey was available. We excluded patients if (i) residing beyond the hospital’s catchment area, defined as home zip code >60 miles from any of the institution’s hospitals, (ii) survey was received by Press-Ganey >60 days following hospital discharge, and/or (iii) the patient refused research authorization or was re-hospitalized prior to returning the survey, to generate a cohort of 40,162 unique patients for analysis.

HCAHPS survey items were on demographics (e.g., race, language) and hospital care items, analyzed both as individual survey questions and as composite measures. Composite measures included communication with nurses (three items: treated with courtesy and respect; listened carefully; explained things in a way you could understand), communication with doctors (three items: treated with courtesy and respect; listened carefully; explained things in a way you could understand), responsiveness of hospital staff (two items: received help when used call button; helped to use bathroom or bedpan), communication about medicines (two items: explained purpose of new medicine; explained side effects of new medicine), discharge information (two items: staff discussed help needed after discharge; received information on symptoms or health problems), and care transition (three items: staff considered needs after discharge; understood how to manage health; clearly understood purpose of medicines); individual items included cleanliness of hospital environment and quietness of hospital environment; and global items included hospital rating and recommend hospital. During the study period, two survey items on pain management changed, and we a priori decided to exclude them. The response to individual items was categorized as favorable (i.e., 'top-box' response if selected ‘always’, score of ‘9 or 10’, ‘yes’, ‘definitely yes’, or ‘strongly agree’) or unfavorable ('non–top-box') response. For composite measures, the overall response (favorable or unfavorable) was determined from responses to the individual items, as described by CMS.[13] A favorable composite measure indicated selection of favorable response to all items in that measure.[11] We defined complete survey responses as answering >50% of items for that composite measure. If a patient had an incomplete composite measure, they were excluded from the analysis for that composite measure but could be included in the analysis for other composite measures.

Patient and Hospitalization Characteristics

Patient and hospitalization characteristics, including the primary outcome, were electronically abstracted from the Mayo Clinic Unified Data Platform, which contains pooled electronic medical record data.

Primary Outcome

The primary outcome was ‘30-day hospital admission’, defined as admission to any of the institution’s hospitals within 30-days after Press-Ganey received the HCAHPS survey.

Statistical Analysis

We compared patient characteristics using Kruskal-Wallis (continuous variables) and Chi-Square (categorical variables) tests. We evaluated the association of survey responses (individual items and composite measures) with odds of 30-day hospital admission using logistic regression: for each outcome, model 1 included age, sex (men, women), race (White, other, unknown), and study site (A–D); model 2 included model 1 covariates plus service line (medical, surgical) and Elixhauser comorbidity index (continuous) in the overall cohort and in subsets stratified by service line (medical or surgical) and by comorbidities (tertiles of Elixhauser comorbidity index). Results were reported as odds ratio (95% confidence interval). Two composite measures (responsiveness of hospital staff; communication about medicines) with branching logic were not reported because sample size was ~75% of overall. We conducted a sensitivity analysis in patients residing ≤30 miles of any of the institution’s hospitals and who returned the survey within 30 days after hospital discharge.

Data were analyzed using SAS^® 9.4 (SAS Institute Inc., Cary, NC) with statistical significance at 2-tailed P<0.05.

Ethics Approval

This study, the Hospital Internal Medicine Patient Reported Outcomes Versus Experience study 1 (Hospital IMPROVE-1), was conducted by the Hospital Experiences to Advance Goals and Outcomes Network (HEXAGON) at Mayo Clinic[14-16] and was deemed ‘Exempt’ by the Mayo Clinic Institutional Review Board.

Results

Baseline Characteristics

Of 40,162 patients (mean age: 68.1 ± 14.0 years) including 19,982 women, 95.1% were of White race (n=38,210/ 40,162) and 98.8% spoke English as the primary language (n=39,692/ 40,162) (table 1). The 30-day hospital admission rate was 4.3% (n=1711/ 40,162). Patients with 30-day hospital admission, compared to those not admitted, were older, more likely to be women, more likely to be discharged from a medical service line, and had a higher Elixhauser comorbidity index (table 1). There were no differences in race, ethnicity, primary language, or study site.

Table 1:

Patient Characteristics Based on 30-Day Hospital Admission^a

Characteristics	30-day hospital admission after receipt of HCAHPS survey by the third-party vendor		P value
	Admitted n=1711	Not admitted n=38,451
Age at admission, years	69.6 (±14.1)	68 (±14.0)	<0.001
Women	792 (46.3)	19 190 (49.9)	0.003
Race			0.54
White	1623 (94.9)	36,587 (95.2)
Black or African American	27 (1.6)	480 (1.2)
American Indian or Alaska Native	2 (0.1)	54 (0.1)
Asian	9 (0.5)	329 (0.9)
Native Hawaiian or Other Pacific Islander	2 (0.1)	31 (0.1)
Other	21 (1.2)	370 (1.0)
Unknownb	27 (1.6)	600 (1.6)
Ethnicityc			0.08
Hispanic or Latino	15 (2.4)	194 (1.5)
Non-Hispanic or Latino	597 (92.1)	12,503 (94.1)
Unknown	36 (5.6)	588 (4.4)
Primary language			0.64
English	1693 (98.9)	37,999 (98.8)
Other	18 (1.1)	452 (1.2)
Service Lined			<0.001
Medical	1076 (62.9)	16,282 (42.3)
Surgical	635 (37.1)	22 169 (57.7)
Elixhauser comorbidity index	6 (4–9)	4 (2–7)	<0.001
Study sitee			0.14
A	280 (16.4)	5641 (14.7)
B	174 (10.2)	4062 (10.6)
C	554 (32.4)	13,198 (34.3)
D	703 (41.1)	15,550 (40.4)
Time from discharge to survey receipt, days	28 (21–34)	29 (23–36)	<0.001

Age reported as mean (± standard deviation). Other data reported as median (interquartile range) and frequency (%).

^aPatients living ≤60 miles of a Mayo Clinic hospital and HCAHPS survey was received by the third-party vendor ≤60 days after hospital discharge. Admission was to any Mayo Clinic hospital ≤30 days after survey was received by the third-party vendor.

^bIncludes unknown race, chose not to disclose race, and missing values.

^cHispanic or Latino included Central American, Cuban, Hispanic or Latino, Mexican, Puerto Rican, South American, and other Spanish culture of origin regardless of race.

^dBased on Medicare Severity Diagnosis Related Groups classification.

^eMayo Clinic Arizona, Mayo Clinic Florida, MCHS, and Mayo Clinic Rochester, randomly labelled A–D. MCHS is a network that includes 17 hospitals in Minnesota and Wisconsin.

Missing values for ethnicity (n=26 229) and Elixhauser comorbidity index (n=1).

P value from Kruskal-Wallis (continuous variables) and Chi-Square (categorical variable) tests.

Abbreviation: HCAHPS Hospital Consumer Assessment of Healthcare Providers and Systems; MCHS Mayo Clinic Health System.

Patient Experience

Patients with 30-day hospital admission, compared to those not admitted, had a lower proportion of favorable responses to 13 of 19 survey items, including all survey items for communication with nursing, communication with doctors and responsiveness of hospital staff (table 2). Of all items, ‘received information on symptoms or health problems’ (discharge information) received the highest proportion of favorable responses (91.1% [admitted] vs 92.1% [not admitted]; P=0.16). The responses to the following 6 survey items did not differ by 30-day hospital admission status: explained purpose of new medicine, explained side effects of new medicine, staff discussed help needed after discharge, received information on symptoms or health problems, staff considered needs after discharge, and quietness of hospital environment. For both global items, the proportion of favorable responses was lower in patients with 30-day hospital admission compared to those not admitted.

Table 2:

Patient Experience Responses Categorized by 30-Day Hospital Admission^a

	30-day hospital admission after receipt of HCAHPS survey by the third-party vendor
	Admitted n=1711	Not admitted n=38,451	P value
	Favorable response, no. (%)
Composite Measures
Communication with Nurses
Treated with courtesy and respect	1513 (89.3)	34,711 (91.2)	0.007
Listened carefully	1346 (79.3)	31,107 (81.8)	0.011
Explained things in a way you could understand	1290 (76.5)	30,210 (79.7)	0.001
Communication with Doctors
Treated with courtesy and respect	1509 (89.0)	34,605 (91.2)	0.003
Listened carefully	1358 (80.3)	31,741 (83.7)	<0.001
Explained things in a way you could understand	1256 (74.5)	29,771 (78.8)	<0.001
Responsiveness of Hospital Staff
Received help when used call button	1061 (67.1)	24,800 (72.2)	<0.001
If needed, helped to use bathroom or bedpanb	791 (69.6)	19,804 (75.7)	<0.001
Communication about Medicines
Explained purpose of new medicineb	903 (78.1)	21,109 (79.8)	0.16
Explained side effects of new medicineb	566 (51.6)	13,081 (52.0)	0.81
Discharge Information
Staff discussed help needed after discharge	1319 (89.2)	30,672 (90.0)	0.32
Received information on symptoms or health problems	1287 (91.1)	30,380 (92.1)	0.16
Care Transition
Staff considered needs after discharge	899 (53.8)	20,955 (55.9)	0.09
Understood how to manage health	899 (53.6)	22,239 (59.0)	<0.001
Clearly understood purpose of medicines	945 (56.8)	22,783 (60.7)	0.002
Individual Items
Cleanliness of Hospital Environment	1289 (76.2)	29,976 (79.2)	0.003
Quietness of Hospital Environment	1004 (59.8)	23 047 (61.1)	0.27
Global Items
Hospital Rating	1345 (79.7)	31,613 (83.4)	<0.001
Recommend Hospital	1379 (82.1)	32,193 (85.2)	<0.001

^aPatients living ≤60 miles of a Mayo Clinic hospital and the HCAHPS survey was received by the third-party vendor ≤60 days after hospital discharge. Admission was to any Mayo Clinic hospital ≤30 days after survey was received by the third-party vendor.

Favorable response indicates the most satisfactory response: always; score of 9 or 10; yes; definitely yes; strongly agree.

Response rate differed across items, in particular, for items with branching logic^b.

P value from Chi-Square test.

Abbreviation: HCAHPS Hospital Consumer Assessment of Healthcare Providers and Systems.

Patient Experience and 30-Day Hospital Admission

In model 1, for the four composite measures, favorable response, compared with unfavorable response, was significantly associated with an 11%–19% lower odds of 30-day hospital admission, except for discharge information, which showed no association (figure 1). Favorable responses to most individual items that constituted the composite measures were associated with significantly lower odds of 30-day hospital admission (online supplemental table 1). For individual items, favorable response for ‘cleanliness of hospital environment’, but not ‘quietness of hospital environment’, was associated with lower odds of 30-day hospital admission. The lowest odds of 30-day hospital admission were observed for favorable response for hospital rating (model 1; adjusted OR, 0.75; 95% CI, 0.66–0.84; P<0.001) and recommend hospital (model 1; adjusted OR, 0.76; 95% CI, 0.67–0.86; P<0.001) (figure 1). Further adjusting model 1 with service line and Elixhauser comorbidity index (model 2) attenuated the odds of 30-day hospital admission for hospital rating (model 2; adjusted OR, 0.88; 95% CI, 0.77–0.99; P=0.039) and eliminated other associations (figure 1 and online supplemental table 1).

Figure 1. Association of patient experience scores (favorable vs. unfavorable) and 30-day hospital admission.

Odds ratios (95% confidence interval) from logistic regression analysis using model 1 (included age, sex, race, and study site) and model 2 (included model 1 covariates plus service line and Elixhauser comorbidity index). Study sites were Mayo Clinic Arizona, Mayo Clinic Florida, MCHS, and Mayo Clinic Rochester. MCHS is a network that includes 17 hospitals in Minnesota and Wisconsin. Patients living ≤60 miles of a Mayo Clinic hospital and the HCAHPS survey was received by the third-party vendor ≤60 days after hospital discharge. Admission was to any Mayo Clinic hospital ≤30 days after survey was received by the third-party vendor.

For composite measures, favorable indicates the most satisfactory response on all items in the composite measure. For individual and global items, favorable indicates the most satisfactory response to the item (score of 9 or 10; definitely yes).

*P<0.05; **P<0.01; ***P<0.001

Abbreviations: CI confidence interval; HCAHPS Hospital Consumer Assessment of Healthcare Providers and Systems; MCHS Mayo Clinic Health System.

In the medical service line, patients with favorable vs unfavorable response for hospital rating had a lower 30-day hospital readmission rate (6.0% vs 7.2%; P=0.007). A significantly lower odds of 30-day hospital admission was associated with favorable vs unfavorable responses for hospital rating (model 2; adjusted OR, 0.81; 95% CI, 0.70–0.94; P=0.007) and recommend hospital (model 2; adjusted OR, 0.83; 95% CI, 0.71–0.97; P=0.021) (figure 2). However, in the surgical service line, patients with favorable vs unfavorable response for hospital rating had similar 30-day hospital admission rates (2.8% vs 2.8%; P=0.91); all composite measures and most survey items showed no association with odds of 30-day hospital admission (figure 2 and online supplemental table 2).

Figure 2. Association of patient experience scores (favorable vs. unfavorable) and 30-day hospital admission categorized by service line.

Odds ratio (95% confidence interval) from logistic regression analysis using model 2 (included age, sex, race, study site, and Elixhauser comorbidity index). Study sites were Mayo Clinic Arizona, Mayo Clinic Florida, MCHS, and Mayo Clinic Rochester. MCHS is a network that includes 17 hospitals in Minnesota and Wisconsin. Patients living ≤60 miles of a Mayo Clinic hospital and the HCAHPS survey was received by the third-party vendor ≤60 days after hospital discharge. Admission was to any Mayo Clinic hospital ≤30 days after survey was received by the third-party vendor.

For composite measures, favorable indicates the most satisfactory response on all items in the composite measure. For individual and global items, favorable indicates the most satisfactory response to the item (score of 9 or 10, definitely yes).

*P<0.05; **P<0.01

When analyzed by tertiles of Elixhauser comorbidity index, favorable vs unfavorable response for hospital rating was associated with lower odds of 30-day hospital admission in the upper tertile (model 2; adjusted OR, 0.79; 95% CI, 0.67–0.92; P=0.003) (figure 3 and online supplemental table 3).

Figure 3. Patient experience scores (favorable vs. unfavorable) and 30-day hospital admission categorized by Elixhauser comorbidity index.

Odds ratio (95% confidence interval) from logistic regression analysis using model 2 (included age, sex, race, study site, and service line). Study sites were Mayo Clinic Arizona, Mayo Clinic Florida, MCHS, and Mayo Clinic Rochester. MCHS is a network that includes 17 hospitals in Minnesota and Wisconsin. Patients living ≤60 miles of a Mayo Clinic hospital and the HCAHPS survey was received by the third-party vendor ≤60 days after hospital discharge. Admission was to any Mayo Clinic hospital ≤30 days after survey was received by the third-party vendor.

Elixhauser comorbidity index tertiles: lower (0–3), middle (4–5), and upper (6–20).

For composite measures, favorable indicates the most satisfactory response on all items in the composite measure. For individual items, favorable indicates the most satisfactory response to the item (score of 9 or 10; definitely yes).

**P<0.01

Abbreviations: CI confidence interval; HCAHPS Hospital Consumer Assessment of Healthcare Providers and Systems; MCHS Mayo Clinic Health System

Sensitivity Analysis

In a sensitivity analysis on patients residing ≤30 miles of any of the institution’s hospitals and who returned the survey within 30 days of hospital discharge (vs ≤60 miles and 60 days in the main analysis), favorable response for hospital rating, compared with unfavorable response, was associated with lower odds of 30-day hospital admission (adjusted OR, 0.78; 95%CI, 0.64–0.95; P=0.013) in the medical service line, consistent with findings from the main analysis (online supplemental table 4).

Discussion

In this multisite analysis of 40,162 patients from an institution’s 20 hospitals across four U.S. states, we evaluated the association of patient experience scores following hospitalization with odds of 30-day hospital admission. A favorable response for hospital rating and recommend hospital was associated with ~20% lower odds of 30-day hospital admission for patients discharged from the medical service line. Similarly, favorable response for hospital rating was associated with ~20% lower odds of 30-day hospital admission for patients in the upper tertile of Elixhauser comorbidity index. To our knowledge, this is the first report on the association of patient experience scores and a novel patient outcome, namely, risk of 30-day hospital admission after third-party vendor receipt of the patient survey. The results show that routine patient experience scores may inform efforts to identify patients at risk for 30-day hospital admission.

In previous studies, favorable responses to care received during hospitalization, always listened, and discharge process were associated with a lower risk of 30-day hospital readmission (i.e., 30 days after hospital discharge).[17,18] A large hospital-level study using the U.S. CMS Hospital Compare database showed that higher overall patient satisfaction and satisfaction with discharge planning in patients with myocardial infarction, heart failure, and pneumonia were associated with a 3%–5% lower risk of 30-day hospital readmission.[7] A cross-sectional study of 30 968 hospitalized patients showed that a favorable response to nurses listened and doctors explained on the HCAHPS survey was associated with 18% lower odds of 30-day hospital readmission.[5] Most studies on outcomes in hospitalized patients have focused on risk of 30-day hospital readmission. However, HCAHPS surveys may be returned up to 90 days following discharge and have the potential to inform risk of other outcomes >30 days after discharge. Therefore, the current study builds on previous work and used a novel outcome to assess the association between patient experience scores and healthcare outcomes in a real-world setting. This outcome is important because it maximizes the use of surveys returned >30 days after discharge and provides a 30-day window of time to implement interventions to potentially reduce risk of 30-day hospital admission. We did not examine the association between patient experience scores and other outcomes (e.g., mortality, emergency department visits), but a recent cross-sectional study suggested that higher online rating by patients of healthcare facilities was associated with lower mortality.[19] Other outcomes following receipt of surveys will be evaluated in future studies.

The literature on the association between patient experience and outcomes is mixed. In an observational study of outpatient visits, denial of patient request (e.g., request for referrals, laboratory tests, or pain medication) was associated with lower satisfaction with the clinician, which could disincentivize cost-conscious care.[20] In contrast, a quality improvement study using Medicare fee for service claims and CAHPS surveys showed no association between low value care (i.e. care not associated with clinical benefit) and improvement in patient experience.[21] Therefore, initiatives to improve patient experience may not necessarily increase healthcare spending or entail a trade off with high value care.[22,23] An additional concern regarding patient experience scores and hospital readmissions is the risk of reverse causality. A single center study of HCAHPS surveys showed that patients who responded to the survey after hospital readmission were more likely to be dissatisfied, suggesting that readmission is predictive of patient experience scores, and not vice versa.[24] In the current study, we eliminated the potential for reverse causality by excluding patients who were readmitted before returning the HCAHPS survey.

HCAHPS summary scores differ across hospitals, service lines, diagnoses, and other case-mix factors.[25-27] Surgical patients, compared with medical patients, tend to have higher summary scores, which have been attributed to streamlined discharge protocols in surgical patients.[11] However, scarce evidence exists on the variation of the association between patient experience and hospital readmissions between service lines within the same institution. Findings from the current study show that higher scores for hospital rating and recommend hospital were significantly associated with lower risk of 30-day hospital admission in medical patients. Higher scores for hospital rating were significantly associated with lower risk of 30-day hospital admission in the upper tertile of Elixhauser comorbidity index. Lower 30-day hospital admission rates in the surgical service line and in the lower two tertiles of Elixhauser comorbidity index may have contributed to the lack of association between hospital rating and 30-day hospital admission.

Describing the drivers of patient experience scores may inform interventions to improve patient experience and outcomes. Several studies reported that patient experience is better predicted by the quality of communication and patients’ perception of their healthcare providers competencies rather than factors such as the hospital environment.[7,28-30] A previous study identified 102 factors associated with different HCAHPS categories.[31] However patient satisfaction with discharge procedures and their communication with nurses and physicians may be key factors in most experience categories, especially for hospital rating.[11,32] A meta-analysis showed that interventions to improve communication were associated with better patient experience and significantly lower risk of hospital readmission.[33] The results from the current study suggest that hospital rating scores may be used to identify patients at risk for 30-day hospital admission following receipt of the survey.

Strengths and Limitations

Our study has potential limitations. Most respondents were of white race and English-speaking, and the significance of these findings in other racial groups and non-English speaking patients requires evaluation. We did not capture hospital admissions outside of the institution’s hospitals. However, the main and sensitivity analyses yielded similar results, bolstering the overall conclusions. Despite the large sample size and the study being conducted at an institution’s hospitals in four U.S. states, the generalizability to non-respondents and other institutions requires evaluation. While HCAHPS survey responses may guide policies and decisions,[34,35] further research is required to evaluate the significance of the nonresponse bias. We did not account for patients’ adherence to discharge recommendations or interim events (e.g., access to primary care, emergency department visits), which may have affected the risk of hospital admission.[36,37] We included surveys received by the third-party vendor from 2–60 days after hospital discharge. However, the potential concern about the broad time window is mitigated by the generally similar results from the sensitivity analysis. Analyses showed that hospital rating was the only global item that remained significantly associated with risk of 30-day hospital admission. Further analysis of the facets of patient experience driving this association needs evaluation. Our study has several strengths. To our knowledge, this is the first analysis of patient experience scores and risk of 30-day hospital admission following third-party vendor receipt of the survey. While the sites were within the same institution, the variability in geographic characteristics and care providers across sites in four U.S. states increases external generalizability but requires evaluation in populations different from the current study’s respondents. We examined the association of patient experience scores and hospital admission based on service line and comorbidity burden, which heretofore, has not been reported. In summary, this study leveraged routinely collected patient experience scores and shows that hospital rating identifies patients at risk for 30-day hospital admission.

Conclusion

In this multisite analysis of 40,162 patients discharged from an institution’s 20 hospitals, we evaluated the association of routinely collected patient experience scores and risk of a novel outcome, namely, 30-day hospital admission following receipt of survey. Favorable hospital rating was associated with lower risk of 30-day hospital admission for medical patients and for patients with a high comorbidity burden. These results identify patients for interventions to reduce risk of 30-day hospital admission.