Prevalence and Predictors of Mortality for Older Adults Referred to a Hospital Avoidance Program

Abstract A cross sectional retrospective data linkage study of older adults discharged from local hospital avoidance program between January 2017 and January 2018 was undertaken (N=286; mean age 80.5 years). The prevalence of death at 3 months, 6 months, 12 months, 18 months and 33 months was calculated. Patient demographic characteristics associated with participant’s risk of mortality at 33 months after discharge was examined using Cox multivariable regression. Patient demographic and health characteristics associated with participant mortality within 12 months of discharge was examined using multivariable logistic regression for patients with complete health characteristic data (n=195). The mortality prevalence was 17% at six months and the cumulative prevalence at one year, 18 months and 33 months post discharge were 24%, 29% and 36% respectively. Patient demographic characteristics associated with participants’ risk of mortality at 33 months after discharge were gender, age and household arrangements. Health and demographic characteristics associated with mortality within 12 months of discharge were lower cognition, increased burden of comorbidity, decreased physical function, a weight less than 55 kilograms, older age and male gender. These results indicate that a significant proportion of people attending a hospital avoidance program are likely to be entering into the final year of their life. This suggests that hospital avoidance programs should routinely identify patients who are likely nearing end of life, and support advance care planning for this patient group.


Introduction
The rate of preventable hospitalizations reflects the performance of healthcare systems in terms of both the quality of acute care and the integration of hospital and community health services. 1 Hospital avoidance programs are implemented in pre-discharge, post-discharge delivered by acute services, post-discharge in the home, or a combination of these. 2 Commonly, older people are the focus of hospital avoidance programs given their relatively high admission and readmission rates, which account for >59% of hospital admissions in the United States. 3 In Australia, people aged ≥65 accounted for 42% of discharges and 48% of patient days during 2016-17. 4 They have a relatively high prevalence of comorbidity, polypharmacy, cognitive impairment, social isolation, low levels of activities of daily living and eventually increased risk of death. A 2015 study of 2350 people aged >70 years over an average of 2.6 years found that excessive polypharmacy (>10 drugs) was an independent predictor of mortality (hazard ratio 1.83). People experiencing frailty and excessive polypharmacy were more than six times more likely to die (hazard ratio 6.30). 5 Hospital avoidance programs are targeted at people with the highest risk of presentations, admissions or readmissions. However, this susceptibility of this population group to death may also need to be actively considered and reflected in their care. 6 This need is reinforced by the equivocal findings of a 2019 systematic review that examined the effectiveness of hospital avoidance programs. 2 Although in the majority of 90 studies included in the review focused on readmission and length of stay outcomes, four studies also reported on mortality. [7][8][9][10] Two of the four studies included patients who were generally at risk of readmission and reported no intervention effect on mortality prevalence. 7,9 The remaining two studies only included patients with heart failure; with one study reporting a significant intervention effect and the other no effect on mortality. 8,10 Reorientation of care to consider susceptibility to death, can also provide an opportunity to discuss future planning, including disclosure of probable life expectancy, completing advance care directives, adoption of a palliative approach aiming to support quality of life in an environment of the person's choosing. 11 Decisions regarding the focus of care provided should be influenced by mortality rate over a 12-month period and those factors that convey a higher risk of death occurring. Consequently, a retrospective data linkage study was undertaken for people attending a hospital avoidance program that aimed to: 1. Mortality prevalence at three, six and 12 months, 18 months and 33 months post discharge from the program; 2. Patient demographic and health characteristics associated with mortality up to 12 months and 33 months postdischarge.

Design and setting
A retrospective data linkage study of older adults discharged from a hospital avoidance program (Healthy at Home) offered by one local health district (LHD) in New South Wales between January 2017 and January 2018.  12 These patients get comprehensive multi-component holistic health and wellbeing assessment, which aims to identify the cause of the person's deterioration, and development of a care plan, which focuses on management of health, psychological and social contributors to deterioration, restoration of function and avoidance of hospital presentation.

Data collection
Patients who attended the avoidance program in the defined time period were identified by the LHD's electronic medical records. A manual audit of the electronic records of these individuals was undertaken. Demographic characteristic data and disease codes were extracted from the electronic medical records by an automated report. The patient data set was linked to the Australian Births Deaths and Marriages database to obtain death data.

Date of death
The date of death (if death occurred) was collected on October 1, 2019 for all patients. This data collection date meant death outcomes were collected up to 33 months post-discharge from the Healthy at Home program.

Health characteristics
Comorbidity. For patients who had been an inpatient or presented to an emergency department between January 1, 2016 and January 31, 2019, their International Statistical Classification of Diseases and Related Health Problems 10th Revision (ICD-10) codes were extracted. The ICD-10 codes were used to calculate the Charlson Comorbidity Index for each patient and a score of ≥5 is considered to be at risk of mortality. 13 The five most prevalent comorbidities were also reported.

Functional ability. Scores on the Timed Up and Go and the
Karnofsky Performance Status scale were extracted to assess functional ability. Timed Up and Go time at ≥12.6 s is associated with fall risk and frailty. 14 The Karnofsky Performance Status describes a patient's functional status. 15 Polypharmacy and anticholinergic burden. The number of regular medications taken by each patient was extracted from the medical record. In addition, all known medications contributing to anticholinergic burden were extracted to calculate the anticholinergic burden score. A total anticholinergic burden score of ≥7 is considered as a high burden score. 16 High anticholinergic burden not only affects cognition, but also reduces appetite, causes functional impairment or frailty and eventually increases mortality. 17 Depression. The 15-item Geriatric Depression Scale score was extracted. The overall sensitivity of the Geriatric Depression Scale is 84.3% and has a specificity of 73.8%. A score of ≥5 is suggestive of depression but a score >10 is almost always associated with depression. 18 Cognition. Weight. Patient weight in kilograms was extracted from the medical record. Owing to the limited data available in the medical record, weight <55 kg was considered an "at-risk" weight.

Demographic descriptors
The following descriptors were collected age, sex, marital status, Aboriginal and/or Torres Strait Islander status, and living arrangements.

Statistical analysis
Statistical analyses were programmed using SAS v9.4 (SAS Institute, Cary, NC, USA). For Aim 1, the sample constituted all patients discharged from the Healthy Home program between January 2017 and January 2018. For Aim 2, the sample constituted only patients for whom all health characteristic data were available (complete cases). For all association analyses, characteristics identified at P < 0.05 were considered statistically significant. Variables were summarized as frequencies with percentages of non-missing observations, means, standard deviations, medians, and minimum and maximum values as appropriate. The anticholinergic burden score was calculated using standard techniques. 23 For each health characteristic, the proportion of patients who were categorized as being at risk was calculated based on accepted cut points or risk levels for established measures or those used by the Healthy at Home clinicians for pathology results (see Table 1 footnote).

Prevalence of mortality post-discharge
For all patients discharged from the Healthy at Home program within the study period, time of death within 3, 6, 12, 18 and 33 months post-discharge was calculated.
Patient demographic characteristics associated with risk of death up to 12 and 33 months post-discharge Cox multivariable regression analysis examined health and demographic characteristics associated with risk of death up to 12 and 33 months post-discharge. For regression analysis, age was treated as a continuous variable. The Timed Up and Go, Geriatric Depression Scale and some pathology outcomes (B12, folate, sodium, creatinine and hemoglobin) were not included as these variables contained the highest missing number of observations. Weight was dichotomized to <55 kg and ≥55 kg. Household arrangements were collapsed into two categories, i.e., "Alone" (lives alone) and "Not alone" (lives with someone). Hazard ratios (HRs), 95% confidence intervals (CIs) and Pvalues were calculated. Sensitivity analyses were conducted to assess how results changed depending on treatment of missing data. Missing observations were imputed using multiple imputation with the full conditional specification method with 100 imputations. All variables hypothesized to have an association with death were included in the sensitivity analysis. This included the Timed Up and Go, Geriatric Depression Scale, Glomerular Filtration Rate, Thyroid Stimulating Hormone, B12, folate, sodium, and haemoglobin. Creatinine was not included as it is contained in the Charlson Index as part of end-stage kidney disease.

Results
During the data collection timeframe, 286 patients were discharged from the Healthy at Home program. Patients were mostly women (56%), lived alone (43%) or with their spouse or partner (39%), and had a mean age of 80.5 years. Patients could have multiple reason for referral to the program. The more prevalent

Health characteristics
The health characteristic of the sample are provided in Table 1.
The prevalence of risk was highest for polypharmacy with 84% of patients taking more than five medications, followed by Timed Up and Go scores (81%), Karnofsky Performance Scale scores (77%), Geriatric Depression Scale scores (52%) and glomerular filtration rate (48%). The five comorbidities with the highest prevalence that contributed to the Charlson Index were congestive heart failure (28%), diabetes with chronic complications (27%), renal disease (27%), chronic obstructive pulmonary disease and other respiratory disease (22%), and dementia (17%).

Prevalence of mortality post-discharge from Healthy at Home program
There were 104 deaths (36%) for the whole patient sample (N = 286) up to 33 months' post-discharge from the Healthy at Home program (see Table 2). This is equivalent to 0.219 deaths per person years follow-up. Nearly half of these patients (n = 48) had died within 6 months of discharge, and 69 (24%) had died within 12 months of discharge from the program.
Patient health and demographic characteristics associated with death up to 12 and 33 months post-discharge At 12 months, the complete case sample consisted of 195 patients and these were included in the regression analysis. Of the complete cases, there were 51 (26%) deaths up to 12 months postdischarge from the Healthy at Home program. Scores on the MMSE, Charlson Comorbidity Index and Karnofsky Performance Scale, and weight and sex were associated with death within 1 year (see Table 3).   program. As at 12 months, scores on the MMSE, Charlson Comorbidity Index and Karnofsky Performance Scale, and weight and sex were associated with death within 33 months (see Table 4). In addition, age, an interaction between household arrangement and time, and albumin were significant in the model. In the sensitivity analysis, the MMSE, the Karnofsky Performance Scale and albumin became insignificant.

Discussion
This study described the mortality of patients attending a hospital avoidance program. It found that 24% of patients had died within 12 months of being discharged from the program. This finding is in line with previous research examining the mortality of patients attending hospital avoidance programs. Dhalla et al. reported a mortality rate of 25.8% at 12 months post-discharge from the inpatient stay for patients deemed to be high risk who were enrolled in a post-discharge virtual ward program. 9 Blum and Gottlieb reported a mortality rate of 19% at 12 months after enrolment to a home tele-monitoring program for heart failure patients. 8  For patients attending hospital avoidance programs, the possible outcomes include improvement, stabilization, deterioration or death. 24 Given the significant prevalence of death for patients discharged from hospital avoidance programs found in this study and others, consideration should be given to including advance care planning for this patient group, particularly for those identified as being more likely to die within 12 months. Although predicting the outcomes for individual patients does remain probabilistic, 24 it is important that efforts be made to do so. Being able to identify patients who have low survival probability would offer a greater opportunity for healthcare professionals, patients and their families to discuss advance care planning. Advance care planning has been indicated to: increase the likelihood that future medical care aligns with patient wishes 25 ; improve patient quality of life 26 ; enable patient management of their personal affairs while they are still able to do so 27 ; reduce stress, anxiety and depression for family members 28 ; and reduce in-hospital deaths while promoting hospice use. 11 A multidimensional geriatric assessment, 29 such as the Multidimensional Prognostic Index are suggested for use in practice as a way to predict outcomes for older people. These assessment tools include comorbidity and the level of physical, cognitive and social impairment. 29 They provide a numerical score that describes the global health of the patient that reflects the risk of negative outcomes, including hospital readmission and death. 29 The use of a comprehensive assessment is supported by the findings of the current study even in the presence of cardiac and respiratory disease. Multiple characteristics were associated with patient mortality. Our study indicated that lower cognition, higher comorbidity, lower functional status, low weight, albumin levels, being older and being male were associated with dying within 12 and/or 33 months post-discharge from the Healthy at Home program. Furthermore, other characteristics that have previously been indicated to be associated with mortality (e.g., anticholinergic burden) were not found to be associated with mortality in this study. 22,30 This may be because of the relatively small sample size included in the current study.
The outcomes of this study should be viewed in light of the study limitations and strengths. Some variables that have previously indicated to be associated with mortality were not able to be included in regression analysis because of missing data. Although the majority of patients in the Healthy at Home program receive all tests and assessments, some components are omitted due to an individualized patient approach. The sample size was relatively small but the prevalence of death outcomes did include all patients discharged from the Healthy at Home over a 1-year period providing a representative sample of that patient population. The findings may not be generalizable as they represent one hospital avoidance program in one local health district of NSW.