DOI QR코드

DOI QR Code

Between-Hospital Variation in All-Cause Mortality for Potentially Avoidable Hospitalizations in Older People

  • Kim, Jae-Hyun (Department of Health Administration, Dankook University College of Health Science) ;
  • Lee, Yunhwan (Department of Preventive Medicine and Public Health, Ajou University School of Medicine)
  • Received : 2018.12.24
  • Accepted : 2019.01.11
  • Published : 2019.06.30

Abstract

Background: Potentially avoidable hospitalizations (PAH) contribute to an increased post-discharge mortality. Methods: To investigate the between-hospital variation and the relationship between all predictors and mortality after discharge among older adults with PAH, we studied 15,186 older patients with PAH in 2,200 hospitals included in the National Health Insurance Service-Senior claims database from 2002 to 2013. Multivariable multilevel logistic regression analyses were performed to analyze the variance at between-hospital for mortality after accounting for differences in patient characteristics. Results: The between-hospital variation in mortality that could be attributed to hospital practice variations were 37.6% at 1-week to 13.9% at 12-month post-discharge, after adjustment for individual patient characteristics and hospital-level factors. Hospital-level factors significantly explained mortality at 3 weeks after discharge. Clinics, compared with general hospitals, demonstrated a 2.75 times higher likelihood of deaths at 3-week post-discharge (p<0.001). Compared with private hospitals, public hospitals exhibited 1.61 times higher odds of 3-week mortality (p=0.01). Conclusion: This study demonstrates considerable between-hospital variations in PAH-related mortality that could be attributed to hospital practices. Monitoring of hospitals to identify practice variations would be warranted to improve the survival of older patients with PAH.

Keywords

INTRODUCTION

Variations in the provision of healthcare services among hospitals have been an important issue of outcomes research on quality of care. This variability in outcomes represents inconsistencies in care, suggesting a considerable room for quality improvement. In the United States, to narrow the variability in the provision of healthcare services for ensuring the quality of care, the National Surgical Quality Improvement Project has allowed comparative assessment of outcomes and developed a set of quality indicators for continuous quality improvement within the Veterans Affairs health system [1].

In a study examining hospital variations, Garnick et al. [2] reported wide variations in 30-day postadmission hospital mortality rates. Hospital rankings by 30- and 180-day mortality rates after admission, however, did not yield markedly different results. Rosenthal et al. [3] also found nearly a three-fold variation across hospitals in both in-hospital and 30-day mortality rates, with substantial variations in discharge practices. In Korea, given the usefulness of mortality as a quality indicator, hospital mortality at discharge has been used for assessing outcomes of hospital practice [4], with 30-day mortality rates after discharge being evaluated in various studies [5,6]. Lee et al. [7], using administrative data, found large variations in hospital standardized mortality ratios among hospitals with more than 700 beds.

Substantial variability in hospital practices suggests that many hospitalizations are potentially avoidable, further contributing to the reduction of adverse outcomes. Identifying factors associated with these variations may further help in delivering patient-centered care. Whether the variation is attributable to patient- or hospital-level factors can contribute to the development of specific strategies for quality improvement. Although potentially avoidable hospitalizations (PAH) are reported to be associated with poor survival rates, the extent to which mortality is attributed to the hospital after discharge is largely unknown. Therefore, in this study, we analyzed national health insurance data collected for older patients with PAH to examine between-hospital variations and the relationship between all predictors and mortality after discharge.

METHODS

1. Data source

We used the National Health Insurance Service–Senior (NHIS– Senior) claim database from 2002 to 2013, released by the National Health Insurance Service in South Korea. To construct the NHIS-Senior cohort, the baseline sample was selected by stratified systematic random sampling to generate a representative sample in 2002. Non-citizens and special purpose employees with an unidentifiable income level were excluded. The sample consisted of 558,147 participants, comprising approximately 10% of people aged 60 years and older in 2002. The cohort was followed annually for 11 years until 2013 unless the participant became noneligible due to due to death or emigration. All databases are linked anonymously using unique encrypted patient codes in accordance with laws on personal information protection. Approval from an ethics committee is not required to analyze encrypted claims data.

2. Potentially avoidable hospitalizations

Based on the Agency for Healthcare Research and Quality indicators [8], PAH which included bacterial pneumonia admission rate, dehydration admission rate, pediatric gastroenteritis admission rate, urinary tract infection admission rate, perforated appendix admission rate, low birth weight rate, angina admission without procedure, congestive heart failure admission rate, hypertension admission rate, adult asthma admission rate, pediatric asthma rate, chronic obstructive pulmonary disease admission rate, uncontrolled diabetes admission rate, diabetes long-term complication admission rate, and rate of lower extremity amputation among patients with diabetes, was defined as incident cases if their first records had a hospital admission based on the International Classification of Disease 10th revision (ICD-10). Both acute and chronic PAH were considered. There were 15,186 patients identified as hospitalized with potentially avoidable conditions.

3. Dependent variable

Our main outcome was all-cause mortality. Data on mortality were collected at the time of 1-week to 12-month discharge. The event of death was counted after the date of first PAH based on ICD-10 code at hospital discharge.

4. Covariates

Both individual- and hospital-level covariates were included. Individual-level covariates included age, sex, residential region, income, Charlson Comorbidity Index (CCI), and route of admission. Age in years was categorized into four groups: ≤69, 70–79, 80–89, and ≥90. Residential region was categorized into metropolitan (Seoul), urban (Daejeon, Daegu, Busan, Incheon, Kwangju, or Ulsan), and rural (otherwise). Income in deciles was categorized into three groups: low (≤3), middle (4–7), and high (8–10). CCI was grouped as scores of 0, 1–2, and ≥3. Route of admission was categorized as emergency, outpatients, and others (decision by a physician included). As a hospital-level variable, hospital type was categorized as general hospital, hospital, and clinics. Ownership of the hospital was categorized as public, corporate, and private. Number of doctors within the hospital was categorized into five groups: ≤49, 50–149, 150 –249, 250–349, and ≥350. Presence of medical devices, such as computed tomography (CT), magnetic resonance imaging (MRI), and positron emission tomography (PET), was also included.

5. Statistical analysis

Chi-square tests were used to compare differences in the 3-week mortality by patient demographics and hospital characteristics. Hospital characteristics were presented as percentages. Because the 15,186 patients were nested within 2,200 hospitals, multilevel models were used to analyze the hierarchically structured data [9]. Multivariable multilevel logistic regression was used to analyze the between-hospital variance for mortality after accounting for differences in patient characteristics. We calculated the intraclass correlation coefficient (ICC) statistic, representing between-group variation as a proportion of all variation for the dependent variable when data are analyzed by groups (e.g., individuals or hospitals) by time of post-discharge death. The proportion of the between-hospital variation for mortality attributable to hospitals was calculated as 1− {(ICC of model accounting for hospital-level factors)÷(ICC of model not accounting for hospital-level factors)}. To account for differences in patient characteristics among hospitals, models were adjusted for the covariates listed previously. All analyses were performed using the SAS software ver. 9.4 (SAS Institute Inc., Cary, NC, USA). Two-sided p-values of less than 0.05 were considered to indicate statistical significance.

RESULTS

1. Sample characteristics

Of the 15,186 hospitalized with potentially avoidable conditions, 390 (2.6%) were deceased at 3-week post-discharge (Table 1). The 3-week mortality rates tended to be higher among hospitals smaller in size, staffed with fewer physicians, funded by the public sector, and not equipped with MRI, CT, or PET scans. Among the individual-level characteristics, older age, male gender, higher comorbidity, and potentially avoidable acute conditions exhibited higher mortality.

 Table 1. Characteristics of participants at first discharge

BGHJBH_2019_v29n2_220_t0001.png 이미지

Table 1. Continued

BGHJBH_2019_v29n2_220_t0002.png 이미지

Of the hospital types, general hospitals accounted for 16.7%, and hospitals 34.1%. Private hospitals (59.8%) accounted for the largest percentage by type of hospital ownership (Table 2).

 Table 2. Characteristics of hospitals at first discharge of participants

BGHJBH_2019_v29n2_220_t0003.png 이미지

2. Between-hospital variation in mortality

Results from the multivariable multilevel models on mortality indicated clustering at the hospital level. The proportion of all variation (ICC) attributable to mortality in models accounting for hospital-level factors steadily increased from 0.060 (95% confidence interval [CI], -0.376 to 0.795) at 1-week post-discharge to 0.146 (95% CI, 0.207 to 0.921) at 3-week, 0.192 (95% CI, 0.581 to 0.985) at 3-month, 0.204 (95% CI, 0.683 to 1.007) at 6 months, and 0.220 (95% CI, 0.783 to 1.067) at 12-month post-discharge, adjusting for patient and hospital characteristics. Between-hospital variations for mortality attributed to the hospital-level factors accounted for 21.6% at 3-week, 17.6% at 3-month, 17.2% at 6-month, and 13.9% at 12-month post-discharge (Table 3).

 Table 3. ICC for variation in mortality by time of post-discharge

BGHJBH_2019_v29n2_220_t0004.png 이미지

3. Relationship between covariates and mortality

Table 4 shows the association between the covariates and 3-week mortality. The mortality risk in the smallest-sized hospital (clinics) was 2.75 times higher compared with that of the general hospital (p<0.001). Public hospitals, compared with private hospitals, demonstrated 1.61 times higher odds of mortality (p=0.01).

 Table 4. Adjusted OR of mortality at 3-week post-discharge

BGHJBH_2019_v29n2_220_t0005.png 이미지

DISCUSSION

In this cohort of older patients with PAH, large between-hospital variations in mortality were observed. Taking into account hospital-level factors reduced the ICC attributable to mortality throughout various time points after discharge, suggesting that characteristics of hospitals explain a significant portion of the difference in mortality rates among those with PAH.

After discharge from hospital, patients are at high risk of readmission and mortality, especially during the first weeks to months after discharge. Few patients, however, have access to multiprofessional disease management programs, which can reduce the risk of adverse outcomes [10]. In addition, factors related to physician’s decision-making, such as perceived low risk, avoidance of patient anxiety, and litigation risk may contribute to the practice variation [11]. Moreover, economic incentives of hospitals or physicians may affect day-to-day clinical decisions [12].

Hospital-level factors significantly explained mortality at 3 weeks after discharge. In contrast, ICC for models including hospital-level factors was not statistically significant at 1- or 2-week mortality post-discharge. It is likely that between-hospital variations at 1-week and 2-week are attributable to individual-level factors, such as disease severity or management. A 3-week post-discharge mortality rate might be a more appropriate quality indicator for assessing the between-hospital variation in survival of PAH patients.

In addition, the results of multivariate analyses revealed higher mortality rates for clinic than for general hospitals and for public hospitals than for private hospitals. As noted, public hospitals are totally controlled by district governments, and this may contribute to difficulties recruiting well-trained physicians, acquiring medical equipment, and the higher mortality rates [13]. Additionally, restrictions of medical equipment for specialized treatments in clinics than general hospital may contribute to higher mortality rates.

The findings of our study have implications for improving survival of elderly patients admitted for PAH. This study shows that the mortality of patients with PAH varied significantly between hospitals, implying that there is room for improvement in survival of elderly patient at the hospital level. In addition to improving survival of older patients admitted to the hospital, interventions tailored for individual hospitals may be necessary to improve patient survival rates. A detailed examination of hospital-level factors is necessary, since such understanding would facilitate implementation of effective interventions to reduce between-hospital variations.

In interpreting our findings, several potential methodological limitations should be considered. First, due to lack of data, our models could not account for some clinical and demographic factors associated with the between-hospital variation in outcomes. Although we found that mortality after discharge was clustered within the hospital, we were unable to determine whether the between-hospital variation resulted from differences in hospital and physician practices or the implementation of quality improvement programs, such as clinical pathways and protocols, affecting patient safety. Second, although we adjusted for severity of illness based on clinical data that were abstracted from the claims database, it is possible that unmeasured severity varied across hospitals or across patients. Finally, it is important that future research examine the applicability of our findings to diverse patient groups, particularly to surgical conditions and conditions associated with either higher or lower post-discharge mortality.

There are considerable between-hospital variations in mortality among older patients with PAH. The between-hospital variation was significantly higher at 3-week mortality after discharge. A close examination and monitoring of hospital and physician practices would be needed to increase the survival of older patients with PAH.

ACKNOWLEDGMENTS

This work was supported by the National Research Foundation of Korea Grant funded by the Korean Government (NRF-2016S1A3A 2925399).

References

  1. Khuri SF, Daley J, Henderson WG. The comparative assessment and improvement of quality of surgical care in the Department of Veterans Affairs. Arch Surg 2002;137(1):20-27. DOI: https://doi.org/10.1001/archsurg.137.1.20.
  2. Garnick DW, DeLong ER, Luft HS. Measuring hospital mortality rates: are 30-day data enough?: Ischemic Heart Disease Patient Outcomes Research Team. Health Serv Res 1995;29(6):679-695.
  3. Rosenthal GE, Baker DW, Norris DG, Way LE, Harper DL, Snow RJ. Relationships between in-hospital and 30-day standardized hospital mortality: implications for profiling hospitals. Health Serv Res 2000;34(7):1449-1468.
  4. Ben-Tovim D, Woodman R, Harrison JE, Pointer S, Hakendorf P, Henley G. Measuring and reporting mortality in hospital patients. Canberra: Australian Institute of Health and Welfare; 2009.
  5. Lee MK. Community acquired pneumonia. Tuberc Respir Dis 2011;70:1-9. DOI: https://doi.org/10.4046/trd.2011.70.1.1.
  6. Jang WM, Eun SJ, Sagong P, Lee CE, Oh MK, Oh J, Kim Y. The change in readmission rate, length of stay and hospital charge after performance reporting of hip hemiarthroplasty. J Prev Med Public Health 2010;43(6):523-534. DOI: https//doi.org/10.3961/jpmph.2010.43.6.523.
  7. Lee EJ, Hwang SH, Lee JA, Kim Y. Variations in the hospital standardized mortality ratios in Korea. J Prev Med Public Health. 2014;47(4):206-215. DOI: https://doi.org/10.3961/jpmph.2014.47.4.206.
  8. Agency for Healthcare Research and Quality. AHRQ quality indicators: guide to prevention quality indicators: hospital admission for ambulatory care sensitive conditions. Rockville (MD): Agency for Healthcare Research and Quality; 2001.
  9. Twisk JW. Applied multilevel analysis: a practical guide. Cambridge: Cambridge University Press; 2006.
  10. Cowie MR. Postdischarge assessment and management of patients with heart failure. Medicographia 2015;37(2):155-162.
  11. Lin GA, Dudley RA, Redberg RF. Cardiologists' use of percutaneous coronary interventions for stable coronary artery disease. Arch Intern Med 2007;167(15):1604-1609. DOI: https://doi.org/10.1001/archinte.167.15.1604.
  12. Gray BH. The new health care for profit: doctors and hospitals in a competitive environment. Washington (DC): National Academy Press; 1983.
  13. Kim SJ, Park EC, Kim TH, Yoo JW, Lee SG. Mortality, length of stay, and inpatient charges for heart failure patients at public versus private hospitals in South Korea. Yonsei Med J 2015;56(3):853-861. DOI: https://doi.org/10.3349/ymj.2015.56.3.853.