Participant retention is vital for longitudinal studies. Home visits may increase retention, but little is known about the subset of patients they benefit.
To evaluate patient-related variables associated with home visits.
In a 5-year, longitudinal, multisite, prospective study of 195 survivors of acute respiratory distress syndrome, in-person assessments were conducted at a research clinic. Home visits were offered to participants who could not attend the clinic. Associations between having a home visit, prior follow-up visit status, and baseline and in-hospital patient variables were evaluated with multivariable, random-intercept logistic regression models. The association between home visits and patients’ posthospital clinical status was evaluated with a subsequent regression model adjusted for these variables.
Participants had a median age of 49 years and were 56% male and 58% White. The following had independent associations with home visits (adjusted odds ratio [95% CI]): age (per year: 1.03 [1.00-1.05]) and immediately preceding visit incomplete (2.46 [1.44-4.19]) or at home (8.24 [4.57-14.86]). After adjustment for prior-visit status and baseline and hospitalization variables, these posthospital patient outcome variables were associated with a subsequent home visit: instrumental activities of daily living (≥ 2 vs < 2 dependencies: 2.32 [1.29-4.17]), EQ-5D utility score (per 0.1-point decrease: 1.15 [1.02-1.30]), and 6-minute walk test (per 10-percentage-point decrease in percent-predicted distance: 1.50 [1.26-1.79]).
Home visits were important for retaining older and more physically impaired study participants, helping reduce selection bias caused by excluding them.
With survivorship after critical illness increasing, understanding patient outcomes after hospital discharge is important.1–3 Studies of these patient outcomes reveal significant impairments for years following hospitalization.4–9 Understanding postdischarge outcomes requires rigorous studies with few patients lost to follow-up.10 Loss to follow-up introduces selection bias11 because participants lost to follow-up may have more severe impairments in functional and socioeconomic status.11,12
Many strategies and resources can help improve retention rates.13,14 For studies that require in-person assessments, one strategy is to conduct home visits for participants who otherwise would not return to complete a study assessment at a research location.15–18 Home visits involve research personnel going to the participant rather than the participant coming to a centralized research location. Home visits require planning and additional resources. Knowing factors associated with home visits can facilitate study design and resource planning. Our objective was to use data from a multisite, 5-year, longitudinal study of survivors of acute respiratory distress syndrome (ARDS) to evaluate factors associated with having a home visit.
Study Population and Design
Patients from 13 intensive care units (ICUs) within 4 teaching hospitals in Baltimore, Maryland, were recruited if they received mechanical ventilation and had a diagnosis of acute lung injury according to the American-European Consensus Conference criteria that were in effect at the time of the study.19,20 Consistent with the more recent Berlin definition, we use the term ARDS rather than acute lung injury.21 Individuals were excluded if more than 96 hours had elapsed between ARDS diagnosis and enrollment or if they had more than 5 days of mechanical ventilation before enrollment, preexisting ARDS at the time of transfer to a study ICU, a life expectancy of less than 6 months based on pre-ARDS comorbidities, a limitation in care at the time of enrollment (other than a sole do-not-resuscitate order), a prior lung resection, an inability to speak or understand English, or no fixed address.
Participants completed a battery of patient-reported and performance-based assessments at each follow-up visit, scheduled for 3, 6, 12, 24, 36, 48, and 60 months after ARDS onset. Loss to follow-up was minimized by using published retention methods, including the following: a letter describing the study with a magnet showing the study logo and contact phone number; visit reminders via phone and letter; free meal vouchers for use at the research clinic; free parking or taxi rides; thank-you letters; annual newsletters and birthday cards; and flexible timing for study assessments (eg, early or late in the day and during the weekend). Home visits were offered to those unable to come to the research clinic.13,20,22–24
Factors Evaluated for Association With Home Visits
The primary outcome and dependent variable for this study was receipt of a home visit, which could occur at each follow-up visit. We captured participant characteristics before hospitalization, during hospitalization, at the time of hospital discharge, and over the course of the 60-month follow-up. Prehospital status variables included residence at home with health care services, participant’s report of chronic fatigue, ability to walk at least 5 minutes without stopping, and score on the EQ-5D. We evaluated the EQ-5D visual analogue scale score (range, 0-100; higher score is better) and the utility score (range, −0.11 to 1.0; higher score is better).25 Additional baseline participant characteristics collected from the medical records included demographics (age, sex, race, employment status, and education level), functional comorbidity index (range, 0-18; higher score is worse),26 history of excessive alcohol use or illicit drug use, and baseline psychiatric comorbidities. Hospitalization factors included durations of mechanical ventilation, ICU stay, and hospital stay. Measures collected at hospital discharge included chronic shortness of breath, independence in all activities of daily living, and discharge location. In addition, starting at the second follow-up visit (6 months after meeting ARDS criteria), the following variables describing the patient at the immediately preceding visit were coded: missing the visit, having an incomplete visit, and having a home visit. We selected these baseline hospital and visit characteristics a priori for this analysis on the basis of existing studies and clinical judgment.
Participant status variables collected at each follow-up visit included independence in all activities of daily living, instrumental activities of daily living (IADLs) (dichotomized as ≥ 2 vs < 2 IADL dependencies),27 chronic shortness of breath, residence location, Hearing Handicap Inventory for Adults–Screening score,28 unemployment due to health, EQ-5D visual analogue scale and utility scores, 36-Item Short Form Health Survey physical component score and mental component score (age- and sex-matched standardized score; mean [SD], 50 ; higher score is better),29 Hospital Anxiety and Depression Scale anxiety and depression subscale scores (range for each, 0-21; lower score is better; scores ≥ 8 indicate substantial symptoms),30 Impact of Event Scale–Revised score (range, 0-4; lower score is better; scores ≥ 1.6 indicate substantial symptoms),31,32 6-minute walk test (percentage of predicted value),33 manual muscle test score (range, 0 to 60; higher score is better),34 hand grip strength (percentage of predicted value),35 forced expiratory volume in 1 second (percentage of predicted value),36 and maximal inspiratory pressure (percentage of predicted value).37 We recorded these status variables for all participants regardless of whether they were assessed at home or at the research clinic. At each follow-up, the reason for not having the visit at the research clinic was noted.
Participants were first categorized according to whether they had ever received a home visit. Across these 2 categories, each patient factor was compared using the Fisher exact test or the Wilcoxon rank sum test, as appropriate.
Using the data from the 6- to 60-month follow-up visits, we separately correlated each baseline, hospital, and prior-visit factor with receipt of a home visit. We used a logistic regression model that also included indicators for follow-up time and a random intercept for each participant to account for the correlation of receiving a home visit over time within each participant.
If a baseline, hospital, or prior-visit factor contained less than 20% missing data and had a significant association in the bivariable regression model (P < .05), it was selected for subsequent multivariable analysis. For participants in whom a factor was measured at admission, during the hospital stay, and at discharge, we included the discharge measurement in the multivariable model. We evaluated the association of a home visit with changes in these selected factors over time by using a second set of these models that included an interaction term between these factors and the indicators for follow-up time. In addition, we used a random-intercept multivariable model including all of the selected factors and indicators for follow-up time to evaluate the association of this set of variables with receipt of a home visit. Finally, we correlated receipt of a home visit at a given follow-up with the participant status variables measured at the same follow-up by using a random-intercept logistic regression model adjusted by the selected baseline, hospital, and prior-visit factors. This model also included an indicator for follow-up time and an indicator for whether the participant status variable was missing at the given follow-up time.
Patient factors before, during, and after hospitalization were evaluated for association with receiving home visits during research follow-up.
A 2-sided P value of less than .05 was used to indicate significance in all tests and models. All statistical analyses were performed with SAS statistical software, version 9.4 (SAS Institute Inc). Ethics approval was obtained from the Johns Hopkins Medicine Institutional Review Board (NA_00025950, NA_00041630), the University of Maryland, Baltimore, Institutional Review Board (HCR-HP-00053863), and the VA Maryland Research and Development Committee (H-26354). Each participant provided written consent.
The cohort in this analysis included 195 of the 196 consenting participants who survived to the 3-month follow-up visit (see Figure). One comatose patient was excluded from the analysis. The median age of participants was 49 years, with 44% female, 58% White, 63% having had no more than high-school education, 39% employed prior to ARDS, 45% having alcohol or drug abuse, and 26% having any psychiatric comorbidity (Table 1).
During the 5 years of follow-up (7 follow-up time points), the percentage of participants who received a home visit decreased from 39% at 3 months to 23% at 60 months (see Figure). For the 125 participants with 1 or more home visit, half of the follow-up visits were conducted at home (median, 50%; interquartile range, 29%-100%). Compared with participants who received no home visits, those with 1 or more home visits were older (51 vs 45 years), were more often White (66% vs 44%), and had a longer median hospital stay (30 days vs 23 days) (all P < .01).
Of the 21 baseline and hospital factors evaluated, 7 were significantly associated with a home visit 6 months to 5 years after ARDS (Table 2). Of the prior-visit factors evaluated, an immediately preceding incomplete visit and an immediately preceding home visit were significantly associated with a subsequent home visit 6 months to 5 years after ARDS. History of alcohol or drug abuse, psychiatric comorbidity, and baseline EQ-5D visual analogue scale and utility scores were not associated with home visits (Table 2).
Sixty-four percent had ever had a home visit. Preceding incomplete follow-up visit and home visit, along with age and physical functioning, were associated with later home visits.
The following 9 factors were included in the multivariable models: age, White race, having no more than a high school education, functional comorbidity index, mechanical ventilation duration, discharge to home with services, shortness of breath at hospital discharge, a preceding incomplete visit, and a preceding home visit (Table 2). In the same models, inclusion of the interaction terms between these factors and follow-up time indicators showed that none of the interaction terms were significant (data not shown), indicating that the relationships between these factors and receipt of a home visit remained relatively constant during the 5 years after ARDS. In a single random-intercept, multivariable logistic regression model including only these factors and follow-up time, the following factors remained significantly associated with a home visit (odds ratio [95% CI]): age (per year: 1.03 [1.00-1.05]), preceding incomplete visit (2.46 [1.44-4.19]), and preceding home visit (8.24 [4.57-14.86]) (Table 2).
In the multivariable models evaluating participant status after hospital discharge during the 6- to 60-month follow-up period, the following measures were associated with a home visit: IADL dependency (≥ 2 vs < 2 dependencies: 2.32 [1.29-4.17)), EQ-5D utility score (per 0.1-point decrease: 1.15 [1.02-1.30]), and 6-minute walk distance (per 10-percentage-point decrease in percent-predicted distance: 1.50 [1.26-1.79]) (Table 3).
Participant illness was the most common reason for not coming to the research clinic overall and was also the most prevalent reason during the early follow-up period of up to 24 months (Table 4). By 36 months, travel distance to the clinic was the most common reason for not coming to the clinic.
In this multisite, prospective cohort study evaluating long-term outcomes in 195 ARDS survivors, baseline and hospital factors associated with a home follow-up visit were age, White race, having no more than a high school education, functional comorbidity index, duration of mechanical ventilation, having been discharged to home with services, and shortness of breath at hospital discharge. After hospital discharge, factors strongly associated with having a home visit at a subsequent follow-up were having a preceding incomplete visit and having a preceding home visit. After adjusting for these significant baseline and follow-up factors, participant dependency in 2 or more IADLs, lower EQ-5D utility score, and a lower percent-predicted for 6-minute walk test at follow-up were independently associated with increased odds of having a home visit.
Retention of participants in longitudinal studies is important to maintain statistical power and reduce threats to internal validity.38,39 Participant and hospital factors associated with a need for a home visit during follow-up point to sicker and older participants. Retaining participants and preventing loss to follow-up should be priorities in longitudinal cohort studies and randomized clinical trials because the confounding prevented by randomization could be lost through exclusion of the very sick. Many of the participants with home visits gave illness as the reason for not attending the clinic visit, and the factors measured at follow-up that increased the odds of a home visit are largely measures of physical function. If home visits were not provided, these participants would not be included and the study results would be biased toward younger, healthier participants.39 History of alcohol or drug abuse, psychiatric comorbidity, and prehospitalization EQ-5D visual analogue scale and utility scores had no association with home visits. We did not collect data on the recency of alcohol or drug abuse or on the level of psychiatric morbidity control. These factors might have been remote or well controlled enough to not affect current behavior in our study. We suspect that baseline function on the EQ-5D instrument was less important than expected because the incident ARDS caused such a profound change in all patients that differences in baseline function were erased.
Had home visits not been done, older and less physically able patients would have missed follow-up visits, likely resulting in biased results.
Participant retention requires good planning, time, and targeted effort.40,41 Therefore, retention methods that could help these participants return for follow-up (eg, paying for transportation for those who live far away from clinic, flexible scheduling, and incentives) should be implemented first.13,42,43 Additionally, providing home visits is a valuable technique to retain participants who would otherwise be lost to follow-up.15,44 As more hospitals and clinicians consider establishing postdischarge ICU recovery clinics for ICU survivors, these findings may provide some preliminary insights into the characteristics of patients who may not be seen in hospital-based follow-up clinics. Some published studies of this type of clinic have shown attendance rates below 50%.45–47
This study has potential limitations. First, participants’ prehospitalization health, physical function, and quality of life status were obtained from retrospective interviews, which may have introduced recall bias. The inability to obtain prospective prehospitalization status is an inherent challenge in studies involving patients with ARDS, given the emergent and unpredictable nature of ARDS onset. Second, the cohort retention efforts employed in this study may differ from those of other studies, which may affect interpretation of the findings. Third, these results may not be generalizable to other patient populations.
Despite these limitations, this study has significant strengths. Notable strengths include our low levels of participant loss to 5-year longitudinal follow-up and our extensive collection of prehospitalization, in-hospital, and longitudinal posthospital data for evaluation as risk factors for loss to follow-up. Additionally, our team was well trained and adhered to the study’s detailed, published retention protocol.24 Finally, the longitudinal design allowed us to examine changes in associations over time, with the finding that associations between evaluated factors and the requirement for home visits remained relatively constant over time.
Critical care survivors, including patients with ARDS, are often burdened with lasting physical and mental health morbidities after hospital discharge. This study demonstrates that age and physical function during follow-up are independently associated with the requirement for home visits as part of a longitudinal follow-up study. This finding highlights the need for studies to have comprehensive retention strategies such as home visits. If our study had not included home visits, older and less physically able participants would have missed follow-up visits and our study results would have been biased.
Lisa Aronson Friedman and Daniel L. Young served as co–first authors and contributed equally to the work. The authors thank all patients who participated in the study and the dedicated research staff who assisted with data collection and management for the study, including Dr Nardos Belayneh, Ms Rachel Evans, Ms Kim Pitner, Dr Abdulla Damluji, Ms Carinda Feild, Ms Thelma Harrington, Dr Praveen Kondreddi, Ms Frances Magliacane, Ms Jennifer McGrain, Ms Stacey Murray, Dr Kim Nguyen, Dr Susanne Prassl, Ms Arabela Sampaio, Ms Kristin Sepulveda, Dr Shabana Shahid, Dr Faisal Siddiqi, and Ms Michelle Silas.
This analysis was supported through a grant from the National Heart, Lung, and Blood Institute (NHLBI R24HL111895). The Improving Care of Acute Lung Injury Patients study, which provided data for this analysis, was supported by the National Institutes of Health (P050HL73994, R01HL088045, and K24HL088551) and the Johns Hopkins Institute for Clinical and Translational Research (UL1 TR 000424-06).
To purchase electronic or print reprints, contact American Association of Critical-Care Nurses, 27071 Aliso Creek Road, Aliso Viejo, CA 92656. Phone, (800) 899-1712 or (949) 362-2050 (ext 532); fax, (949) 362-2049; email, firstname.lastname@example.org.