Work

Risk factors for hospital readmission of patients with heart failure: A cohort study

Year: 2020

Type: article

Abstract: INTRODUCTION Heart failure (HF) is a worldwide medical health problem that contributes to the world’s economic burden for chronic disease management. HF is the most common and leading cause of admission and readmission among elderly.[12] There is a wide variation in the reported readmission rates across HF populations, facility settings, and follow-up periods. Studies have reported 30-day, 90-day, and 1-year readmission rates between 18% and 32%, 21% and 47%, and 50%, respectively.[3456789] The surge of unplanned 30-day all-cause admission and readmissions for HF results in a total expenditure exceeding US$30 billion in the United States alone.[9] Satish et al.[10] suggested that screening of acute hospitalized patients at risk of adverse outcomes of hospitalization, including unplanned readmissions, could be achieved by using clinical electronic record-abstracted geriatric targeting criteria. Consequently, it is important to identify patients who are at risk of early unplanned readmission due to HF. The state-wide of discharge data set from Pennsylvania in the United States shows that comorbidities and sociodemographic factors, including male sex, age, black race, medicare coverage, and prolonged length of stay, were associated with increased risk of readmission in patients with HF.[11] A study analyzed 2279 consecutively hospitalized older adults with decompensated HF across 15 hospitals in the United States. This study reported that home health care use was inversely associated with early readmission. Demographic and cardiovascular risk factors and geriatric syndromes were associated with early readmission. Discharge to home health care may reduce early readmission rates in these patients.[12] Studies have been conducted with the aim of producing hospital readmission risk assessment model, formulating questionnaires based on those models and applying them in practice to enable identification of patients. These studies display a great variation in their methodology and consequent reported results. Hospital readmission logistic regression model needs to incorporate the influence of scheduled visits, unscheduled visits, renal status, adverse drug reactions, abnormal laboratory results, and changes in medication regimen. Recently, there are prediction algorithms to identify patients with HF at high risk of readmission or death post-hospital discharge. The Gulf acute HF registry (Gulf CARE) in five countries has described the clinical characteristics, management, and outcomes of acute HF patients with high readmission rates.[13] The analysis from multicenter registry, the Gulf Registry of Acute Coronary Events (Gulf RACE), revealed that HF is observed in about one in five patients with acute coronary syndrome in the United Arab Emirates (UAE) with increased in-hospital mortality. The independent predictors of in-hospital HF in multivariable logistic regression were age, hyperlipidemia, diabetes, prior coronary artery disease, and heart rate.[14] OBJECTIVE The objective of the current study was to investigate and delineate the risk factors, which may be implicated in putting a patient at greater risk of readmission due to HF. ETHICS APPROVAL The study has obtained ethical approval from the research and ethics committee at the hospital. Invited patients who fulfilled the criteria have consented for the access to their clinical data and for the interview part. MATERIALS AND METHODS This was a cohort study conducted at a tertiary hospital in Al Ain city in UAE. The data were collected over a 6-month duration from January 2015 to July 2015. The study population was drawn from the hospital registry and included all patients with HF who were admitted to the hospital during the study period. The sample size was calculated based on the annual admission in the HF hospital registry. The participating subjects (N = 170) were recruited consecutively over the stated given time period. The hospital has a registry for patients with HF. This registry was used to select a total of 170 patients diagnosed with HF (systolic and diastolic HF with and without left ventricular dysfunction), who were admitted sequentially to the tertiary hospital for the management of their HF during the study period. The inclusion criteria were confirmed diagnosis of HF (systolic or diastolic with or without left ventricular dysfunction), both gender, and currently admitted to the hospital due to HF as the main reason for admission. There were no exclusion criteria. Patient information (medical, medication, demographic, sociodemographic, and socioeconomic) was collected from their clinical profile using a data collection instrument previously developed by Al Deagi et al.[15] In addition, each patient was interviewed (by one specified clinical pharmacist who is the main study investigator) during their hospital stay using a structured questionnaire previously developed by Miller[16] to obtain further patient-specific information. Information obtained from the clinical profile was clarified if necessary during the patient interviews.[16] All data collected were entered into SPSS to facilitate the development of a risk factor model for readmission. Development of risk factor model The procedures used to identify the risk factors for readmission and to build a risk factor model were based on the validated procedure published by Hosmer and Lemeshow.[17,18] The cutoff point for risk assessment was set at 0.5, that is, above this point, a patient was at high risk of one or more non-elective unplanned readmissions over a 12-month period. Statistical analysis Descriptive statistics was used for variables with means and standard deviation (mean SD). We have used univariate and multivariate logistic regression analyses on structured steps based on the published data by Hosmer and Lemshow.[1718] Model building for study population Univariate statistical analysis The application of chi-squared analyses indicated that demographic, medical, and social factors were significantly related (P < 0.1) to one or more hospital admissions in the 12 months before the current admission. Multivariate analysis Using stepwise backward logistic regression analysis, any variables whose univariable test had a P value of less than 0.1 was a candidate for multivariate modeling. The total number of variables with P value less than 0.1 from the chi-squared tests (likelihood ratio tests statistic) was 30 variables. With reference to the total number of patients in the study population (N = 170) and as per the recommendations of Hosmer and Lemshow,[17] the rule of 10 was followed, which suggests a final predictive model with 10 variables (i.e., one-tenth the smaller group of patients). Entering of the significant variables (either from patient clinical data or patient interview) independently into stepwise backward elimination logistic regression analysis with model entry set at P = 0.2, and model removal set at P = 0.2,[18] yielded two separate preliminary models for clinical profile and interview data. Repeating this procedure for further elimination with entry and removal values of P = 0.15, P = 0.01, and P = 0.05 yielded final models for both the patient clinical profile data and the patient interview data. The number of significant variables from the resulting models (the clinical profile and review models) was therefore 14. A further regression was performed to allow the two individual models to be combined with model entry and removal set this time at P = 0.02. A final predictive model for unplanned readmission with 10 variables was produced. When the cutoff point for risk assessment was set at 0.5, that is, above this point, a patient is considered to be at high risk of one or more non-elective readmissions 12 months post-discharge. The Hosmer and Lemeshow goodness-of-fit test gave a chi-squared value of 4.3601 (8degrees of freedom, P = 0.8233); therefore, the model can be regarded as an acceptable fit to the data used. Analysis of pair-wise interactions between variables using a logistic regression entry procedure, with P set at 0.05, illustrated that the interactions did not produce significant P values for log-likelihood ratio tests for final risk model. To check that all variables were independent, the φ coefficient was calculated; all phi coefficients generated from each pair of variables were <0.09; therefore, the variables in the final model can be regarded as independent. RESULTS Characteristics of the study population A total of 170 patients with HF were included in the study; 84 (49.4%) of the patients were males and 86 (50.6%) were females. Thirty-nine males (60.9%) and 25 females (39.1%) were aged less than 65 years, while 45 males (42.5%) and 61 females (57.5%) were aged 65 years or older. Regarding the level of education, 47 (48%) males and 51 females (52%) have received elementary education only, while 37 males (51.4%) and 35 females (48.6%) have received more than elementary education. A total of 34 males (57.6%) and 25 females (42.4%) were still working, while 50 males (45.0%) and 61 females (55.0%) were retired. The mean serum creatinine value was 1.8 mg/dL (SD 0.3) for the readmitted patients versus 1.1 mg/dL (SD 0.2) of one-time admitted patients during the index year. The number of rescheduled visits to the emergency room was calculated based on annual visits without being admitted apart from current readmission index. Risk factors delineated by univariate and multivariate logistic regressions With reference to the total number of patients in the study population (N = 170 patients), the number of patients with more than one admission was 74 (43.5%), and the number of patients with only one admission (de novo admission) was 96 (56.5%). The mean length of stay during the present admission for all patients was 10.3 days (SD 1.2) and at the time of the study the mean number of cardiovascular medications used was 5.5 (SD 0.8). Review of the clinical profile revealed that 58.1% of the patients had a history of ischemic heart disease, 59.5% had hypertension, and 64.7% had diabetes mellitus. The most frequently prescribed medications received by the study patients included β-blockers, diuretics (thiazide and loop diuretic), mineralocorticoids (spironolactone), angiotensin-converting enzyme (ACE) inhibitors, angiotensin receptor blockers (ARBs), nitrates, aspirin, statin, and warfarin. Most of these cardiovascular medications were used in two or less daily doses (69.4%) while 30.6% were used in more than two daily doses regimen [Table 1].Table 1: Sociodemographic data for patients with heart failureThe majority of the study population (93%) were on first-line therapy with β-blockers, ACE inhibitors/ARBs, and furosemide, while some (25%) were on the second-line therapy with mineralocorticoid receptor antagonists (spironolactone). The types of HF included patients with diastolic HF (56%) with preserved left ventricular systolic function, that is, ejection fraction of more than or equal to 50% (Heart Failure with preserved Ejection Fraction, HFpEF), and those with HF with reduced ejection fraction of less than or equal to 40% (Heart Failure with reduced Ejection Fraction, HFrEF) (systolic HF; 44%). The final clinical profile model for unplanned readmission included seven significantly statistical variables (P < 0.05). The specificity of the model was 63.5%, the sensitivity was 83.3%, and the overall accuracy was 74.7% [Table 2]. The final model derived from the patient interview data included seven significant variables. The specificity of the model was 62.2%, the sensitivity was 79.2%, and the overall accuracy was 71.8% [Table 3].Table 2: The clinical profile data of patients with heart failureTable 3: The patient interview dataThe final logistic regression model for unplanned readmission of patients with HF at Al Ain hospital, UAE contains seven factors: four or more prescribed medications, dosing regimen: more than two dosing times/day, poor knowledge of prescribed medications, diabetes mellitus, patient complaining of edema or swelling in the extremities, patient bound to home as a result of HF disease, and being prescribed diuretics. The seven factors were presented with respective variable coefficient B, odds ratio (OR), 95% confidence level, and P values [Table 4].Table 4: The final logistic regression model for unplanned readmission of patients with heart failure at Al Ain hospital, United Arab EmiratesThe specificity of the final developed model was 82.2%, the sensitivity was 74.3%, and the overall accuracy was 72.9% [Table 5].Table 5: Observed and predicted values, sensitivities, and specificities from the logistic regression model for unplanned readmission of patients with heart failureDISCUSSIONS The current study model developed for risk factors of readmission to hospital in patients with HF confers diverse population which may have contributed to versatile predictors. Therefore, local population characteristics should be examined to develop a model that best fit to the specific population rather than emulating any internationally predicted model. We have reviewed 170 patients, readmitted to hospital for the management of HF. Readmission could have been avoided in patients if there had been better assessments, if rehabilitation had been more adequate, if discharge had been more carefully planned, if potential nonadherence with medications and diet had been identified, and if patients had been instructed to seek medical attention when symptoms occur. Much attention has, therefore, been given to identifying patients who are at risk of early unplanned readmissions. The present study yielded a 10-factor risk model for hospital readmission of HF patients in the UAE. The risk factors included the number of prescribed medications (four or more), more than a twice daily dosing regimen, poor knowledge of prescribed medications, having diabetes mellitus, having renal impairment (reduced creatinine clearance), patient complaining of edema and swelling in the extremities, prescribed thiazide diuretic, unscheduled visits to emergency room, and bound to home as a result of HF. Modifiable and nonmodifiable clinical factors Number of prescribed medications The logistic regression model revealed that patients who were taking four or more prescribed medications were about four times more likely to be readmitted to hospital over a 12-month period. This finding was consistent with previous reports, which indicated that an increase in the number of medications being received by a patient leads to increased admission. Dosing (more than a twice daily dosing regimen) The readmission rate in patients who had a dosing regimen, which required them to take prescribed medications, more than twice per day was 55.8%. The logistic regression model revealed that such patients were approximately two times more likely to be readmitted when compared with patients who required less frequent drug administration. While multiple dosing drug regimens are often required for optimal management of chronic diseases, the goals of the therapy are often undermined by poor patient adherence to the prescribed regimen. Studies from western and Middle Eastern countries indicate that HF is an increasingly frequent reason for hospital admission. Better survival rates post-myocardial infarction, coupled with an aging population, are reasons for this trend.[19] The risk model suggests that, where possible, dosing regimens should be simplified to allow twice daily dosing, in agreement with previously published data.[20] Knowledge of prescribed medication The readmission rate in those patients who were assessed for poor knowledge of their prescribed medications (interview data) was 46.5%. In the final model, this increased their risk of readmission by a factor of 4.2. A multidisciplinary intervention study on rates of readmission of patients with HF revealed that readmission could be reduced by comprehensive education of patients and their families, a prescribed diet, social service consultation, a review of medications, and intensive follow-up. Further work by horn and colleagues, highlighted the need for patient education within a comprehensive management program for HF patients.[21] With regards to this particular risk factor, the healthcare professionals including clinical pharmacist has a key role to play. Enormous studies have shown that improved outcomes of drug therapy can be achieved via intensive structured education and counseling,[2223] and the present study lends support to these findings that clinical pharmacist should ensure that HF patients receive medication counseling both at discharge and at outpatient clinics. Edema and swelling in the extremities Edema and swelling in the extremities are sign of poorly controlled HF and therefore it was not surprising that patients who complained about edema were more likely to be re-hospitalized (OR = 2.6). The findings support the importance of monitoring weight changes in patients with HF.[24] Patients bound to home as a result of their HF It has been shown in two previous studies that bed rest and limited activity are detrimental to exercise tolerance and aerobic capacity while exercise training improves vasodilatation and muscle oxidative capacity in patients with HF.[2526] Being housebound was shown in the present study to increase risk of rehospitalization by a factor of 2.8. Being housebound may have both social support and disease severity components, which were not “teased out” in the present study. Nonetheless it is an easy question to ask patients and if a patient is found to be housebound, the intensity of patient education and monitoring should be increased to help avoid early rehospitalization. Prescribed thiazide diuretic The readmission rate in those patients who had been prescribed a thiazide diuretic was 46.6% of readmitted patients. The logistic regression model indicated that patients who have received a prescribed thiazide diuretic were over three times more likely to be readmitted. The adverse effects of thiazide diuretics therapy include complications such as hypokalemia and azotemia. Overuse of diuretics in patients with diastolic dysfunction can also lead to hypo-perfusion of vital organs; therefore, a renal electrolyte panel should be checked after adjusting the thiazide diuretic dose. Renal impairment The model has shown that renal impairment in patients with HF is a risk factor for readmission to the hospital. The degree of renal impairment as measured by creatinine clearance reflects the increased risk with reduced creatinine clearance values. Unscheduled visits to the emergency room The increase in annual number of unscheduled visits to the emergency room for patients with HF has implicated increased risk of readmission due to HF. diabetes The readmission rate of patients with diabetes was the risk model indicated that having diabetes increased the risk of readmission within 12 months by a factor of studies in the as as the the of diabetes as a significant risk factor for HF. This risk factor is not it was that attention should be given to patients with HF who also have diabetes to help ensure that both are controlled This was important in with of diabetes such as study A study of one patients with HF in four of the United States has shown readmission rates of and for the two and The study reported that the were associated with readmission risk after for than 65 years, male gender, race, than or drug renal chronic diabetes, and and electrolyte In the most important finding of the current study to the of medications as predictors of readmission in the developed model for patients with HF. This has on adherence to medications, and knowledge about medications for HF. Therefore, studies should incorporate medications, and in multivariate models for prediction of risk factors for readmission of patients with HF. research should incorporate the clinical disease such as and the Heart of the of patients of weight of patients, and emergency early as factors to be considered in risk of readmission. and The most important of the current study was the developed risk factor model that can be by healthcare professionals to early readmission of patient with HF. The sample size of the current study may some on the of the risk factor model. The of the developed model greater attention in with population characteristics in patients with HF. factors could be as for disease risk did the need for improved patient education about prescribed medications, and the that can result from diuretic It has been observed that patients with diastolic HF were less hospitalized than of this did not in the final developed model. It has been reported that to HF is for in the current this has not been and it may have the admission of study patients. It has been that the use of for patients reduce the heart was a factor to reduce patient admissions for HF. in the current there were patients on which may have the final model as factor for readmissions of a sample of patients admitted to tertiary hospital in the as a result of the factors were to significantly increase the risk of multiple hospital admissions within a 12-month four or more prescribed chronic medications, more than a twice daily dosing regimen, poor knowledge of prescribed medications, of diabetes mellitus, renal (reduced creatinine clearance), with edema and swelling of extremities, and being housebound and prescribed The final predictive model for unplanned readmission of patients with HF has specificity of 82.2%, sensitivity of 74.3%, and the overall accuracy was The developed risk model can be by health professionals of population characteristics to early readmission of patient with HF. on practice early readmission in patients with HF. patients with HF at risk of readmission. discharge for patients with HF. scheduled clinical the use of on discharge medication the clinical role in patients with HF. support and of There are no of We are to the of at the hospital for their in the access to the heart failure patient registry.

Source: Journal of Pharmacy And Bioallied Sciences

DOI: 10.4103/jpbs.jpbs_323_20