POST-ACUTE HOME CARE AND HOSPITAL READMISSION OF ELDERLY PATIENTS WITH CONGESTIVE HEART FAILURE

Elderly patients often receive home care from formal service providers after inpatient hospitalization (Caro & Blank, 1988; Kane, 1994; Kane & Kane, 1987). Elderly patients' use of home care services has grown at an unprecedented rate. Between 1988 and 1994 the proportion of Medicare beneficiaries receiving home care services nearly doubled, and the average number of visits per user almost tripled (Moon, 1996). However, elderly patients receive most of the needed care from family members, relatives, or friends. These informal caregivers help elderly patients with functional needs and participate in the elderly patients' medical care (Kane & Kane). Despite the involvement of informal caregivers and formal service providers, researchers have not paid sufficient attention to the effects of informal and formal home care services on elderly patients' outcomes during the post-acute period (Kane, 1996).

The purpose of this study was to examine the independent and joint effects of informal and formal service use on hospital readmission. This study is important because few researchers have tested the effects of informal and formal home care services on hospital readmission, especially during the post-acute period. The findings will expand our understanding of the interface of informal and formal services and their effects on elderly patient outcomes in post-acute care. The findings may also help policymakers and service providers develop policies and programs that are more responsive to the needs of elderly patients and that may reduce hospital readmission.

LITERATURE REVIEW

Although they vary depending on sample criteria and length of observation, hospital readmission rates for elderly patients with congestive heart failure (CHF) are high (Burns & Nichols, 1991; Kane, 1994). After reviewing more than 17,000 CHF patients' Medicare records, Krumholz and colleagues (1997) noted that over a six-month period 44 percent of patients experienced at least one readmission. Vinson and colleagues (1990) found that the readmission rate for a group of patients ages 70 or older was 47 percent over 90 days. Kane (1994) reported a 24 percent readmission rate in six weeks.

Researchers have suggested that hospital readmission is related to alterable factors such as compliance with medical treatment, seeking medical attention, adequate discharge plans, or well-designed and implemented home health care services (Benack, 1964; Berkman & Abrams, 1986; Proctor, Morrow-Howell, Li, & Dore, 2000; Vinson, Rich, Sperry, Shah, & McNamara, 1990). Through these measures, it has been estimated that 6 percent to 53 percent of hospital readmissions may have been preventable (Proctor et al., 2000; Rich et al., 1995; Vinson et al., 1990). Home care services during the post-acute period are a recognized strategy for delaying or preventing hospital readmission for CHF patients (Berkman & Abrams, 1986; Martens & Mellor, 1997). Medical home health services may stabilize the recovery process, and supportive services may compensate for functional dependency (Kane & Kane, 1987).

Post-acute home care is often shared among informal caregivers and formal service providers. A study showed that up to 93 percent and 63 percent of elderly patients received informal and formal services, respectively (Kane, 1994). Informal caregivers help elderly people with activities of daily living (ADLs), such as feeding, bathing, dressing, toileting, and getting in or out of bed, as well as instrumental activities of daily living (IADLs), such as housekeeping, shopping, transportation, administering medication, or handling finances. Informal caregivers are also actively involved in patients' medical care (Kane, 1996; Lough, 1996; Silliman, Shelah, & Amina, 1996). A qualitative study on post-acute home care for elderly patients with CHF revealed that informal caregivers contacted physicians, verified medication regimens, helped with grocery shopping, and incorporated prescribed diets into family meals (Lough).

However, the effects of informal services on hospital readmission during the post-acute period have not been tested directly (Kane, 1996). Using proxy variables such as marital status or living arrangements, a few researchers have suggested that informal assistance may help prevent or delay hospital readmission. They have argued that informal caregivers assist elderly patients with their needs and encourage them to cope with conditions, which may contribute to recovery (Berkman & Abrams, 1986; Burns & Nichols, 1991). However, due to the differences in sampling criteria, length of study, or proxies of informal support, findings from these studies have been inconclusive. Fethke and associates (1986) studied the hospital readmission patterns of a group of elderly patients for 12 months and showed that marital status was not related to hospital readmission rates at six weeks after discharge, but was related to hospital readmission rates at between six and 12 months after discharge. Berkman and associates (1991) followed patients for three months after discharge and found that married patients were less likely to experience hospital readmission. However, Burns and Nichols revealed that elderly individuals' marital status and living arrangements did not affect hospital readmission 60 days after discharge.

Previous researchers have mainly focused on the role of formal services such as hospital treatment, follow-ups by nurses and physicians, discharge plans, and home health care. Martens and Mellor (1997) found that patients who received home nursing services that monitored patients' conditions, medication compliance, and nutrition were less likely to experience hospital readmission than those who did not receive the services in 90 days. Focusing on post-acute care, Rich and colleagues (1995) developed a nurse-directed, multidisciplinary intervention that included education about CHF and its treatment, discharge planning, frequent home visits, and telephone follow-up. Compared with those in the control group who were provided the usual care, elderly patients in the intervention group had lower readmission rates, fewer days of hospital stay, and delayed first readmission at 90 days after discharge.

West and associates (1997) studied a physiciansupervised, nurse-mediated, home-based intervention with an emphasis on compliance with medication and diet, education, and regular telephone contacts for six months. Results from pre- and postintervention comparisons revealed that patients' hospital readmission and doctor office visits decreased and patients' functional status and health improved. Kornowski and associates (1995) developed an intensive home-care service package including self-monitoring of heart conditions, skilled-nursing care, and services from other professionals such as physiotherapists. Comparing the number of hospital readmissions one year before and after the intervention, the program showed a 62 percent reduction in hospital readmission (3.2 to 1.5 times per year). However, none of these studies discussed the role of informal caregivers.

Earlier studies expanded our understanding on the issue of hospital readmission for elderly patients. However, several important limitations in these studies need to be addressed. First, the interventions are effective in reducing hospital readmission, but they are rarely available in most communities. We do not know whether post-acute home health services used by many elderly CHF patients can produce similar outcomes.

Second, informal caregivers are the primary source of post-acute care, but researchers have not directed adequate attention to the services they provide. Although some researchers use elderly patients' marital status or living arrangements as a proxy for informal service use, those factors reflect neither the availability of informal caregivers nor the types or levels of informal services (Penrod, Kane, Finch, & Kane, 1998).

Third, many elderly patients use both informal and formal services after discharge from hospitals, but no researcher has studied the joint effects of informal and formal service use on hospital readmission. We do not know whether the integration of formal services with informal services produces more desirable outcomes than the use of either formal or informal services alone in reducing hospital readmission. This study focuses on informal and formal home health care during the postacute period and examines their effects on hospital readmission.

CONCEPTUAL MODELS

This study is guided by task-specific theory developed by Litwak (1985). According to task-specific theory, an optimal service outcome is generated through the match between service function and task specificity. Litwak and associates (1994) suggested that formal service providers may be more effective in performing the tasks requiring high technology or labor division (for example, nursing care), and informal caregivers may be more effective in performing tasks requiring proximity, commitment, or affection (for example, preparing meals). A combination of informal care and formal services may accommodate elderly peoples' diverse needs and optimize service outcomes.

Based on task-specific theory, this study examined the joint and independent effects of informal and formal services on hospital readmission. Joint effects of informal and formal services were tested through the variable joint use of informal and formal services. This variable has two categories: use of either informal or formal services and use of both informal and formal services. The independent effects of informal and formal services were tested through the level of informal service use and the level of formal service use.

In addition to informal and formal service use, other factors have been identified in relation to hospital readmission: the elderly persons' demographics, severity of CHF, functional and health status, compliance with medications and diet, and informal caregivers' ability to care (Berkman & Abrams, 1986; Berkman, Millar, Holmes, & Bonander, 1991; Burns et al., 1997; Krumholz et al., 1997; Rich et al., 1995; Vinson et al., 1990). Krumholz and colleagues found that male gender, severity of CHF, poor physical health, and prior hospitalizations were related to more incidences of hospital readmission. To isolate the relationship between informal and formal service use and hospital readmission, the variables were tested and controlled.

The research addressed the following three questions:

1. Controlling for other factors, what is the relationship between informal service use and hospital readmission?

2. Controlling for other factors, what is the relationship between formal service use and hospital readmission?

3. Controlling for other factors, informal service use, and formal service use, what is the relationship between joint informal and formal service use and hospital readmission?

METHOD

Sampling Procedures and Data Collection

This analysis was based on a data set collected for the project Adequacy of Home Care Plans for Chronically Ill Elderly. The elderly patients included in the project were initially hospitalized and subsequently discharged from a large mid-western urban teaching hospital and were 65 years of age or older, diagnosed with CHF, discharged to a home setting, and served by a discharge planner. Following the sampling criteria, discharge planners screened the patients referred by physicians or nurses. Over the study period, discharge planners identified 362 patients, and 296 patients (82 percent) agreed to participate. Excluding the patients (56) who died in the hospital or were discharged to nursing homes, in-hospital data were collected from 240 patients.

The sample used for this analysis was a subsample of the parent project and included 199 participants. Between the index hospital discharge and the two-week follow-up, 31 patients were readmitted to hospitals and five dropped out of the study. These elderly patients were excluded from the analysis because research suggested that readmission during the first two weeks of discharge might result from a premature hospital discharge (Ashton et al., 1987). In addition, five elderly patients who did not use any informal or formal services were excluded.

Elderly patients' demographic information and caregiver limitations were collected through a review of medical records and in-hospital interviews. Information on informal and formal service use, functional status, compliance with medication, self-rated health, severity of CHF, and length of CHF history was collected through in-home and telephone interviews at the two-week follow-up. The telephone interview questions were pretested, and test-retest reliability, through Kappa statistic, was at least .80 (Proctor, Morrow-Howell, & Chadiha, 1993). After discharge, elderly patients were monitored for 14 weeks, and hospital readmission information was collected through both medical record review and follow-up interviews. Data were collected between June 1989 and December 1991.

Measurement

The dependent variable was hospital readmission. Hospital readmission was defined as the first, if any, readmission occurring within two to 14 weeks after hospital discharge. Hospital readmission was measured dichotomously according to whether patients experienced hospital readmission from two to 14 weeks post-discharge (readmitted versusnonreadmitted) and continuously through the number of days between hospital discharge and the first hospital readmission, if any. The time interval was selected for the following reasons: First, although the post-acute period was not clearly defined, most of the researchers used six to 12 weeks after discharge as a cut-off (Berkman et al., 1991; Kane, 1994; Rich et al., 1995). Second, Medicare's home health benefit typically ends at 10 to 12 weeks after discharge (Proctor et al., 1993). Third, two weeks of time lag may allow informal and formal services to be put in place, and then time is needed for services to have an effect (Proctor et al., 1993).

The major independent variables were informal service use, formal service use, and joint use of informal and formal services. Informal service use was defined as the use of nursing care, ADL assistance, and IADL assistance provided by unpaid helpers. Formal service use was defined as the use of nursing care, ADL assistance, and IADL assistance provided by paid helpers. ADLs or IADLs refer to the activities elderly people need to perform while living in the community independently (Fillenbaum, 1988). ADLs include getting in or out of bed, walking, toileting, bathing, grooming, dressing, and eating; IADLs include meal preparation, shopping, money management, travel, medication administration, and housekeeping (Fillenbaum). Nursing care focused on checking blood pressure and soundings of heart and lungs, drawing blood, giving injections, and providing catheter care.

To assess the levels of service use, respondents were asked whether they needed assistance in ADLs, IADLs, or nursing care at the interview. If respondents answered yes, they were asked how many episodes of formal or informal service they received in each area in the past week. The level of formal service use was the total number of formal service episodes elderly patients received in each dependent area in the past week. Similarly, the level of informal service use was the total number of informal service episodes elderly patients received in each of the dependent areas in the past week.

Joint use of informal and formal services was defined as the configuration of service provision. It had two categories: use of both informal and formal services (joint use =1) and use of either informal or formal services (joint use = 0).

The control factors tested in this study were elderly persons' demographics, severity of CHF, functional and health status, compliance with medical treatment, and informal caregivers' ability to care. Age was the elderly persons' chronological years of life. Gender (man or woman), race (black or white), marital status (married or not married), and living arrangements (living alone or living with someone) were dichotomous measures. Education attainment was measured on a seven-level scale, ranging from seventh grade or less = 1 to some graduate school = 7 (Hollinshead, 1957). Medicaid status was also a dichotomous variable (yes or no). Socioeconomic status was measured by a twofactor index developed by Hollingshead that consists of education and occupation. This index is reliable with a Cronbach's alpha of .70.

Severity of CHF was measured on a four-level scale developed by the New York Heart Association Classification (Fisher, 1972). Based on a clinical assessment of physical activity, nurses made a judgment on the severity of patients' heart condition. This classification had four categories, ranging from least severe = 1 to most severe = 4 (Fisher). Number of chronic conditions was measured by the sum of chronic conditions that patients identified from the Chronic Condition Checklist that includes 32 chronic conditions such as liver disease, kidney disease, and cancer or leukemia (Fillenbaum, 1988). Length of CHF history was measured by the question of when CHF was diagnosed.

Functional status of elderly individuals was defined as the extent to which the elderly individuals were able to perform ADLs and IADLs. In each of the seven areas of ADLs and the six areas of IADLs, respondents were asked to report their levels of dependency as total dependence = 0, some independence - 1, or total independence = 2. Functional status was indicated by a sum score of levels of dependency in each area, ranging from O to 26. A lower score indicated higher levels of dependence. In this study, Cronbach's alphas for the ADL scale and the IADL scale were .88 and .85, respectively. Self-reported health was measured by the question, "How would you rate your health today?" Respondents were asked to rate their health on a four-level scale, ranging from excellent to poor. The self-reported health measure is reliable and widely used in health services research (Mossey & Shapiro, 1982). Compliance with medication was measured by two questions: (1) "Do you take your medication at the same time every day?" and (2) "During the past week, have you missed any of your tablets?" Patients complied with medication if they answered yes to the first question and no to the second question. Otherwise, patients did not comply.

Caregiver limitations were assessed by yes or no answers to five questions: (1) whether caregiving tasks outweighed caregiver's physical ability; (2) whether the tasks were too complex; (3) whether the caregiver was unreliable; (4) whether the caregiver had too many other duties; and (5) whether the caregiver was on a temporary basis. The caregivers had limitations if there was a yes answer to any of these questions. Otherwise, they had no limitations.

Statistical Analysis

Univariate analyses including percentages, means, and standard deviations were used to describe sample characteristics, including participants' age, gender, race, marital status, living arrangements, functional dependency, cognitive status, severity of CHF, hospital readmission, and informal and formal service use. Because of the small sample size, parsimonious, multivariate models were built through bivariate analyses including t test and chisquare. Only the factors related to hospital readmission at a significance level of 0.10 were retained in the final models. Among the 16 variables tested, elderly persons' educational attainment (t = 1.91, p = .06), marital status (?^sup 2^(1, N = 198) = 4.0, p = .04), caregiver limitations (?^sup 2^(1, N = 199) = 6.1, p = .01), functional status (t = 1.95, p = .05), compliance with medication (?^sup 2^( 1, N = 195) = 4.76, p = .03), self-rated health (t = 1.91, p = .06), number of chronic conditions (t = -2.43, p = .01 ), and length of CHF history (t= -2.10, p = .03) met the criteria and were retained for further testing.

Regression diagnostic procedures were performed to assess multicollinearity and indicated collinearity between functional status and informal service use. When residualized informal service use was entered into the final models, the variance inflation factor for both variables was reduced to an acceptable level (Fox, 1991). As a result, residualized informal service use was used in all three Cox regression models.

Cox proportional regression analysis was used to test the relationship between hospital readmission and informal service use, formal service use, and joint use of informal and formal services. This analysis was selected because of its function of dealing with censoring data or incomplete observations of a given event (Allison, 1995). By the end of the 14-week follow-up, 65 percent of the elderly patients were censored because they had no hospital readmission.

To obtain an accurate estimation, influential data points were examined. The deviance residuals for each independent variable were in the acceptable range, thus there was no indication of problematic outliers (Allison, 1995). In addition, the constant proportionality of the hazard ratio, an assumption of the Cox proportional model, was examined. Insignificant interactions between each independent variable and the timing of hospital readmission indicated that the assumption of the Cox regression model was met in this study. A p value of less than 0.05 was considered to indicate statistical significance.

RESULTS

Sample Characteristics

The average age of respondents was 77.5 years, and ages ranged from 65 to 98 (Table 1). More than half of the patients were women, and slightly fewer than two-thirds were unmarried. After discharge, three-quarters of the respondents returned to live with their spouses, relatives, or friends.

Given the location of the study hospital and the study diagnosis, a little more than half of the respondents were African American. About 55 percent of the participants came from a lower socioeconomic background, and 55 percent had less than high school education. All the respondents were Medicare beneficiaries, 20 percent were enrolled in Medicaid, and 56 percent were covered by private health care plans.

Respondents experienced severe and multiple chronic conditions. All respondents had a diagnosis of CHF, and three-quarters of them had been diagnosed with CHF for more than three years. Forty-five percent of the patients scored 3 or 4 on the New York Heart Association Classification (Fisher, 1972), which indicated severe limitations on patients' activities. Respondents also experienced other chronic health problems such as arthritis, asthma, and kidney disease. On average, they experienced five such health problems, and some of them reported up to 11 conditions.

Given their heart problems and other health conditions, respondents experienced high levels of functional limitations. At two weeks after discharge, 70 percent and 96 percent of the elderly respondents needed assistance in at least one area of ADLs or IADLs, respectively.

Eighty-three percent of the respondents had multiple hospitalizations during the preceding two years. The average length of hospital stay was 14 days. From the two-week to 14-week follow-up, 35 percent of the elderly respondents had at least one hospital readmission.

Informal and Formal Service Use

Two weeks after discharge, 87 percent of the respondents received care from informal caregivers, 89 percent received care from formal service providers. Moreover, 9 percent relied on formal services only, 11 percent relied on informal services only, and 80 percent received services from both informal and formal sources.

Formal and informal services appeared to have different focuses. The top five most commonly used formal services were nursing care (85 percent), assistance in bathing (32 percent), housekeeping (27 percent), meal preparation (19 percent), and grooming (11 percent). This use of services reflected the availability of Medicare home care services and the Meals-on-Wheels programs. In contrast, informal service use seemed to focus on IADLs, and the top five most commonly used informal services were shopping (78 percent), housekeeping (66 percent), meal preparation (53 percent), administering medication (43 percent), and money management (41 percent). However, a large proportion of the respondents used both informal and formal services in meal preparation and housekeeping.

Respondents were more likely to use informal services for help with their functional limitations and formal services for medical care. For example, 85 percent of the respondents used formal services, and only 6 percent used informal caregivers for nursing care. However, in the areas of ADLs, 86 percent of the respondents used informal caregivers, and 39 percent used formal providers; in the areas of IADLs, 46 percent used informal caregivers and 34 percent used formal providers. On average, elderly patients used 40 episodes of informal services and nine episodes of formal services per week.

Relationship between Service Use and Hospital Readmission

The relationship between informal and formal service use and hospital readmission was examined in a Cox multivariate regression model. Due to missing data in some independent variables, 192 cases were analyzed. The model was significant in predicting hospital readmission [?^sup 2^(10, N = 192) = 25.63, p = .004]. After controlling for other variables, formal service use was not significantly related to hospital readmission (risk ratio = .89, CI: .68 - 1.18, p = .97). In the same model, informal service was also not significantly related to hospital readmission (risk ratio = 1.004, CI: .78 - 1.29, p = .42) (see Table 2).

To examine its unique contribution, joint use of services was added to the previous model. In the new model, informal service use and formal service use were also controlled. The model was significant in predicting hospital readmission [?^sup 2^(11, N = 192) = 25.77, p = .007]. However, joint use of services was not significantly related to hospital readmission (risk ratio = .72, CI: .31 - 1.67, p = .45) (Table3).

In both models, hospital readmission was consistently predicted by length of CHF history and compliance with medication. Elderly patients who experienced a longer history of CHF were more likely to be rehospitalized, but elderly patients who complied with medication were less likely to be rehospitalized.

In addition, elderly patients' marital status and education attainment were marginally related to hospital readmission. However, the number of chronic conditions, self-rated health, functional status, and informal and formal service use were not significantly related to hospital readmission.

DISCUSSION AND IMPLICATIONS

This study examined relationships of informal service use, formal service use, and joint use of informal and formal services to hospital readmission during the post-acute period. At discharge many elderly patients may not be fully recovered from the acute conditions and often present high levels of service needs (Goldberg & Estes, 1990). As a result, the proportion of respondents using informal and formal services and the levels of informal and formal service use can be high. Consistent with other studies, findings from this study showed that the vast majority of elderly patients were cared for by informal caregivers (Kane, 1994) and that informal caregivers provided much higher levels of service than formal service providers (Tennstedt & McKinlay, 1989).

The relationship between service use and hospital readmission was examined through Cox regression analyses. Findings indicated that these models significantly predicted hospital readmission. However, there was no evidence to suggest that informal service use, formal service use, or joint use of informal and formal service reduced hospital readmission.

Several factors may have contributed to the insignificant findings. First, the theoretical link between service use and outcomes has not been well understood (Hawes & Kane, 1991). Researchers have proposed that use of health services may improve patient's health and well-being (Andersen, 1995), but often they fail to specify or rationalize the relationship. In the study of home health services, researchers rarely justify the relationship between the services provided and the outcome anticipated. We often do not know whether use of home health services is directly related to service outcomes, such as improvement in functioning or well-being, or through other intermediate outcomes, such as meeting dependency or medical care needs (Penrod et al., 1998). It is important to examine how intermediate outcomes may affect service outcomes in future studies.

Second, this study was a secondary data analysis, and we were limited to the variables that were already collected. For example, the joint use of services independent variable was created to reflect the conceptualization proposed in task-specific theory. This categorical variable may not capture the complexity of how informal and formal services were integrated. Another independent variable, use of informal or formal services, was measured by number of service episodes per week. This measure provided no information on the quality of services or whether services met the needs of elderly patients. Moreover, informal and formal services in the study focused on nursing care, ADLs assistance, and IADLs assistance and did not include services provided by physicians and other professionals, which could have resulted in an underestimation of formal service use. In addition, the dependent variable was defined as any hospital readmission. We did not know whether the readmission was due to CHF or to other unrelated conditions.

Third, the variations in the service delivery process may trivialize the effects of informal and formal services on hospital readmission. Post-acute home health care service providers vary from highly trained professionals to paraprofessionals to informal caregivers. Meanwhile, service delivery and quality may vary from agency to agency. It is challenging to control these variations in a study with a relatively small sample size (Hawes & Kane, 1991). Given the available data, we were only able to separate informal and formal services and identify and control some other confounding factors in the models, which were not adequate to minimize the threats to internal validity.

Fourth, the measure used to correct collinearity may have minimized the effects of informal services on hospital readmission. Consistent with other studies, findings from this study show that functional status of elderly people is highly correlated with levels of informal services. It is understandable that informal service provision is dependent on the functional needs of elderly people (Kemper, 1992; Kane & Kane, 1987). Residual informal service use, a part of informal service use that is not correlated with functional status, may not be able to produce the same effect on hospital readmission as informal service use does.

Fifth, the study participants were recruited from a large, urban teaching hospital, which may limit the generalizability of the study. Last, data were collected between 1989 and 1991. Certain issues during this period may or may not be relevant to current health care.

The study findings confirm that compliance with medication is critical in reducing the risk of hospital readmission. From a practical perspective, hospital discharge planners and home health care professionals need to emphasize compliance with medication in the discharge and home health care plan. This could be done by providing large-print and easy-to-read booklets or video programs that explain the treatment for CHF, the importance of compliance with medication, and the side effects of medication. Such items may enhance patients' and their family members' understanding of CHF treatment. Compliance with medication could also be emphasized in home visits or other contacts with elderly patients and their family members.

In addition, hospital discharge planners and home health professionals should carefully assess patients' financial situations, because the high costs of prescription drugs can create financial burdens, especially for low-income Medicare beneficiaries. Findings of this study show that 22 percent of the respondents had financial difficulties that impeded their ability to purchase medications. The expansion of Medicare coverage to include prescription drugs may ease patients' financial burden and enhance compliance, which may reduce hospital readmission. Hospital social workers and home health agencies need to link patients who lack financial resources to appropriate public assistance programs such as Medicaid or pharmacy assistance programs.

Several resources have been allocated to postacute home health care. Being cared for in one's own home is a universal value for elderly patients and their families. Although this study was unable to demonstrate the relationship between service use and outcomes, research efforts on quality of and balance between informal and formal services should continue (Fischer & Eustis, 1994). More specifically, the interface between informal and formal service can be articulated at different levels, for example, type and volume of services or service change over time. Researchers need to improve the quality of measurement for service use because inconsistency and poor quality in measurement may be responsible for divergent findings. Finally, because the relationship between service use and hospital readmission is weak, replication studies in large samples are needed. HSW