Development of Electronic Transition Tools for Home Health Care

Final Contract Report

Summary

This report details the process of total redesign and automation of the form used to initiate home health services (i.e., the CMS Form 485)—referred to as the "e-485" in this report—and a Web-based electronic tool to facilitate care coordination among hospital physicians, home health nurses, and primary care physicians—referred to as "e-transitions" in this report.

The critical motivation for development of these tools was to improve communication and increase efficiency in the referral process from inpatient to home health care by:

  • Promoting physician involvement in discharge planning.
  • Increasing the amount and accuracy of information available to providers at the time of patient transitions.
  • Promoting the adoption of evidence based practices.

The study team evaluated the impact of the redesigned CMS Form 485 on the transition process and outcomes of heart failure and other patients discharged from the hospital to home care from various perspectives (provider, patient, and family members). Details on this evaluation are also presented.

The electronic referral system of the CMS Form 485 (e-485) that draws on the data available in an electronic health record to generate comprehensive referral orders for patients requiring home health care services was successfully implemented during the project. The e-485 is actively used by clinicians in the geriatrics and HIV units at Weill Cornell Medical College to generate referral orders to the Visiting Nurse Service of New York as well as to multiple home health care agencies.

Results

Processes of care. The e-485 significantly improved a number of process measures of quality of care (compared with handwritten orders) including:

  • Completeness of information on the referral form relating to:
    • Medication orders (e-485: 95% vs. control: 26%, p<0.001).
    • Allergy orders (64% vs. 17%, p<0.001).
  • Inclusion of important information on the referral form such as:
    • Therapeutic goals (88% vs. 66%, p<0.001).
    • Functional status limitations (91% vs. 26%, p<0.001).
    • Mental status (98% vs. 7%, p<0.001).
    • Skilled nursing orders (76% vs. 37%, p<0.001).
  • Evidence-based orders for heart failure patients (post: 45% vs. pre: 11%).

Outcomes of care. Using logistic regression and controlling for such characteristics as age (under 65 vs. over 65), gender, and comorbidities, the e-485 was associated with lower hospital use in the 30 days after referral to home health. There was a net reduction in the odds of emergency department (ED) visits by 72% (odds ratio: 0.28, p=0.012) and a net reduction in the odds of ED visits or hospital readmission by 62% (0.38, p=0.010).

E-transitions. During the 6-week pilot study, e-transitions Web pages for 46 patients were accessed over 1,300 times by referring physicians and nurses. The overall response to the e-transitions pilot testing was positive both for its efficiency and effectiveness in enhancing care coordination. More work at integrating this tool into day-to-day practice is needed for the full impact to be achieved. Benefits included the following:

  • Timely alerting of physicians when patients are being discharged.
  • Facilitation of non-urgent care.
  • Ease in tracking recent orders and assuring they are signed.

Improvements needed include enhanced feasibility for urgent communications, and ability to verify that communication is received and read. Implementation concerns included the ability to check E-mail daily and participants' willingness to take on the additional responsibilities such as periodically checking E-mail, accessing the Web site on daily basis, and taking responsibility for screening E-mails to look for urgent messages.

Conclusion

The project piloted approaches to improving coordination of care during transition from the hospital or outpatient setting to home health care. The e-485 demonstrated important increases in quality of care and was associated with a lower use of ED and hospital services in the subsequent 30 days. Furthermore, a Web-based interactive referral system ("e-transitions"), which was also developed for this project, allows physicians and home health care professionals to exchange referral information, actively manage patients, and exchange data from patients under home health care management.

Return to Contents

 

Background and Purpose

Coordinating the care of persons with chronic conditions is challenging, particularly in the transition from hospital to home care. Decisionmaking is shared by a number of providers in different settings who often lack access to key clinical information. All too often the result is inadequate care planning at hospital discharge and inadequate monitoring of changing conditions after discharge. Communications between physicians and home health nurses often are brief and are rarely timely. Multiple handoffs increase the risk of error. The consequences can be unnecessary rehospitalization, premature decline in functioning, and preventable deaths.

Traditionally, working with home health agencies has not been a priority of physicians and other primary care providers. In the hospital, the patient's admitting physician performs an assessment, develops a plan of care, and writes a series of orders. Nurses, therapists, and other professionals perform separate assessments and together work as a team to implement a plan of care.

Communications may be imperfect, but they occur in real time. The physician is not just involved but is primarily responsible for the patient's progress. The return home of a hospitalized patient, on the other hand, is typically viewed as an end to an episode of care rather than as another admission requiring new diagnoses, goals, and orders.

The need to promote a complete set of orders at hospital discharge and a collaborative plan of care in the home is clear. Ehrenberg1 and colleagues, for example, have described a need for more supportive guidelines to improve home health care for older patients with heart failure. In an effort to reduce unplanned hospitalizations, health care providers and home health care agencies have developed new strategies for transition and care management.2

Harnessing technological advances is one method of improving coordination among health care providers. The advent of electronic health records (EHRs) and new information technologies makes it possible to develop prototype tools so that care is not negatively affected when patients move to long-term care settings from the hospital. Electronic tools that facilitate access to clinical information and communication among clinicians in the hospital and home care setting have the potential to improve transitions and reduce preventable adverse outcomes.

The project described in this report builds on a recently completed Integrated Delivery System Research Network (IDSRN) project that examined the feasibility of using electronic tools to improve communication and care monitoring between members of a transition team at New York Presbyterian (NYPH) Hospital and the Visiting Nurse Service of New York (VNSNY, a home health agency) for patients with heart failure. The aim was to encourage physicians to provide relevant clinical input into the care planning process and monitor care after hospital discharge. The earlier IDSRN project:

  • Expanded the content and reformatted the CMS Form 485 ("Home Health Certification and Plan of Care"a).
  • Pilot tested the new, enhanced form.
  • Assessed the next steps for improving communication electronically.

The purpose of the current project was to evaluate the impact of the e-485 on the completeness of patient referral information and provision of evidence-based home health care. An additional goal was to develop and pilot test a Web-based tool (e-transitions). This tool:

  • Provides access to the information (that is, electronic communication of the e-485).
  • Allows alerts to be designed when appropriate.
  • Gives access to and input from community-based clinicians, who typically do not have access to electronic clinical records.
  • Facilitates subsequent communications between physicians and home health agency staff, as well as with patients and their families.

Using this system, physicians are able to sign plans of care, revisions, and new orders electronically, and nurses and physicians may leave notes for one another on a secure Web site. (This system also enables those physicians who lack EHR capability to participate by manually entering data if they choose and by accessing the system via any Web browser.)

Return to Contents 

Evaluation of One-Way Discharge Form (e-485) 

Development of e-485

Overview and rationale. The rationale for and development of the e-485 has been discussed by Siegler and colleagues.3-4 The CMS 485 is the primary document by which various health providers communicate and develop a plan of care for patients. Completion of the CMS 485 usually necessitates multiple phone calls and faxes from hospital discharge planners,5 social workers, and/or ambulatory care nurses. Not only does its layout lack logical flow and discourage review of orders, it serves as a poor conduit of information because the physician usually signs the form well after home care has been initiated.

Moreover, although physicians must sign home care orders, they do not typically have an active role in preparing the CMS-485. Lack of physician involvement can lead to suboptimal utilization of services and to errors, the most dangerous of which involve medication errors.6-9 The e-485 design was created with the goal of addressing these deficits and as well as the following additional aims:

  • Improve quality by increasing physician involvement and enhancing legibility and communication.
  • Encourage the ordering physician to develop a plan of care that would follow from diagnoses, functional status, and goals
  • Promote cost-effective, appropriate utilization of skilled services by providing evidence-based order sets
  • Be easy to read, naturally guiding the physician to review the data before signing

Appendix A compares the handwritten (CMS 485) and electronic (e-485) forms; a sample e-485 form is presented in Appendix B. The latter is designed so that the commonly used options and prompts are built into the form. For example, the physician notification section has options for specifying when the physician should be notified of changes in the patient's condition. Thus, the e-485's logical format encourages comprehensiveness, accuracy, and improved efficiency.

The e-485 was implemented in the Wright Center on Aging and in the inpatient geriatrics unit in 2004 and was available at the same time in the General Internal Medicine practice at Weill Cornell-NYPH with additional training and efforts to encourage the use of the e-485 occurring in 2005. The e-485 form was also being used in the HIV specialty clinic outpatient electronic health record system in early fall of 2005.

The e-485 continues to be used by the geriatrics clinic and the Center for Special Services at Weill Cornell-NYPH. The e-485 form is being used to refer patients to other home health agencies in addition to VNSNY. The geriatrics clinic and HIV clinic continue to regularly utilize this system to refer patients to home health agencies.

Modifications to e-485 in response to user suggestions. Real time use has led to a number of changes. Researchers at VNSNY asked the manager of the admissions unit at this large, nonprofit agency whether her staff had experienced any problems with the form. The response was that certain orders did not appear clearly over the fax machine and font sizes were increased to improve legibility. Wound care orders were modified and more demographic information was added in response to other comments. The HIV Service requested order sets for indwelling intravenous catheters. The e-485 was modified in response to these user requests.

Processing of e-485s by VNSNY Central Admissions Unit. E-485 referrals from Weill Cornell-NYPH are submitted by fax to the VNSNY Central Admissions Unit (CAU). Staff at the CAU are responsible for typing all demographic and plan of care information into the VNSNY mainframe system for viewing by clinicians in the field and billing personnel. The e-485 includes orders that were not traditionally on the CMS-485. Some of the disease-specific orders (e.g., home health orders for patients with diabetes) were based on existing VNSNY order sets; the VNSNY electronic health record includes fields for these orders, which facilitated CAU data entry.

However, other order sets (e.g., those for patients with heart failure) were new to VNSNY, and there were no fields in the VNSNY electronic health record designated for these orders. The current procedure for recording information at referral for which there are no existing fields for the data in the VNSNY electronic health record is to type in orders as free text in the nursing note screen, and CAU staff followed this process. The results presented below suggest that the expanded e-485 home health orders were most often entered into the mainframe by CAU staff where fields already existed for the orders.

After CAU compiles the orders, a possible home care recipient is assigned to a home care nursing team based on the ZIP Code of residence. Up until this point the possible recipient is in the "pre-admit" phase of home care admission. A patient cannot be admitted into VNSNY home care until a home care nurse on the designated team makes an initial visit to evaluate services needed. During the "pre-admit" phase, the plan of care information on the mainframe is subject to many revisions by CAU, billing, and home care nursing staff. The plan of care at the point that the patient is formally admitted may be substantially different from the original referral sent by Weill Cornell-NYPH. Throughout the subsequent episode of care, VNSNY staff frequently updates the plan of care to reflect changing needs of the patient in the home.

Return to Contents 

Analysis of e-485

As noted above, the revised e-485 form was used to refer patients to home care by the geriatrics and HIV clinics and by the inpatient geriatrics unit at Weill Cornell-NYPH during the enrollment period for the evaluation. E485s were sent to VNSNY as well as other local home health agencies. The sample discussed here, however, is limited to patients referred from Weill Cornell-NYPH to VNSNY because the project was not able to capture data on patients referred to other home health care agencies using the e-485.

From December 1, 2004, through July 31, 2006, e-485 referrals were faxed to the VNSNY Central Admissions Unit. CAU staff copied the e-485 referrals and sent them to the Research Center for use in the evaluation. The research team received a total of 130 referral forms.

Three factors contributed to the smaller-than-anticipated size of the e-485 sample, which limits the types of analyses that can be conducted, especially subgroup analyses of specific disease groups. These factors are:

  • Many referrals from NYPH do not go to VNSNY, and e-485s sent to other agencies are not included in the study sample.
  • The software system was changed during the course of the study, which excluded a number of users from access to the e-485.
  • There was a lack of uptake of the e-485 by the general medicine and surgery departments.

Two study designs were used to evaluate the impact of the e-485 on patient care:

  • A concurrent control group study compared handwritten CMS 485 referrals with to e-485 referrals to home health care.
  • A pre-post study compared a control group consisting of patients referred for home health care in the 12 months prior to initiation of the e-485 with patients referred using the e-485.

Concurrent control group study. The control group consisted of 108 patients referred for home health care from Weill Cornell- NYPH using the traditional handwritten CMS 485. The intervention group consisted of 130 patients referred for home health care from Weill Cornell- NYPH using the e-485.

Data were abstracted from the referral forms (CMS 485 or e-485) and from the Weill Cornell-NYPH medical record by trained abstractors. Abstracted data were entered into an electronic database for analysis. No individual personal identifiers were collected, and only aggregated data are presented here. Descriptive statistics and regression modeling were conducted using SPSS v. 12.0. Demographic information and comorbidities are shown in Table 1. 

Table 1. Characteristics of control/handwritten group (n=108) and intervention/e-485 group (n=130) study subjects

CharacteristicPercent in control group (n)Percent in intervention group (n)
Female60.2 (65)71.5 (93)
Average age, in years7582
Comorbidities:    
HIV/AIDS0.9 (1)11.0 (14)
Diabetes16.7 (19)17.7 (23)
Congestive heart failure9.3 (10)17.7 (23)
Hypertension46.3 (50)63.8 (83)
Pressure sores16.7 (18)17.7 (23)
Cancer39.8 (43)20.5 (27)
Cardiovascular accident5.6 (6)12.6 (16)

Table 2 shows that completeness of important clinical and prognostic information was significantly greater for the e-485 group than for the handwritten group. Generally, the criteria for determining "presence" or "completeness" versus "absence" or "incompleteness" were straightforward—either information was there or it was not there for the category in question. In some instances, as with diagnoses and medications, the evaluation involved comparing the information involved in each category with other pieces of information available on the referral form, as well as by making a judgment based on the overall picture of each patient. For instance, if the patient had a medication listed for diabetes and diabetes was not listed as a diagnosis, the diagnosis would be considered incomplete. 

Table 2. Percentage of information that is present and complete on the referral form

CharacteristicPercent handwrittenPercent e-485P
Demographics96.399.20.383
Mental status7.498.4<0.001
Diagnoses73.497.7<0.001
Prognosis85.295.4<0.001
Goals65.787.7<0.001
Functional limitations25.990.8<0.001
Medications80.795.4<0.001
Allergies17.463.8<0.001
Physician notification17.335.4<0.001
Skilled nursing orders36.776.2<0.001

1Physician triggers include any care orders that pertain to specific instances in which the physician should be notified, such as in cases of significant change in blood pressure, pulse, or weight.

Table 3 compares the e-485 and handwritten referrals on four measures: completeness of orders for diabetes and wound care and rates of hospital readmission and ED visits within 30 days of referral. There were no differences in wound care orders, but analysis revealed the e-485 was associated with a significant impact on the rate of hospital readmission and on the rate of ED visits within the 30-day period after referral to home health care. 

Table 3. Comparison of handwritten and e-485 referrals on four measures

 Percent handwrittenPercent e-485P
Diabetes care (n=42)33.366.70.077
Wound care (n=41)43.256.80.480
Hospital readmission119.48.70.037
Emergency department visit119.46.30.007

1Within 30 days of referral to home health care.

Analysis was also done to ascertain any change in the usage of ACE inhibitors and beta-blockers among those patients who were identified as having a diagnosis of congestive heart failure in the comparison sample. No differences were found; but when comparing the intervention to control groups, the poor power resulting from small sample sizes (only 33 patients had congestive heart failure) prevented detecting any differences in the use of these medications.

A logistic regression model was constructed to predict ED visits or hospital readmissions. Univariate analyses were run to determine any correlation between these two dependent variables and the following independent variables: age (under 65 vs. over 65), gender, referral type (e-485 or handwritten), and comorbidities (cancer, cardiovascular accident, congestive heart failure, diabetes, hypertension, and HIV/AIDS). The univariate analyses showed that the following variables were significant or near significant at the p<0.20 level of significance:

  1. Emergency department visit—Variables included: age (p=0.188); cardiovascular accident (p=0.143); referral type (p=0.007).
  2. Hospital readmission—Variables included: age (p=0.005); referral type (p=0.037).
  3. Either emergency department visit or hospital readmission—Variables included: age (p=0.030); referral type (p=0.012).

Based on these univariate tests, the regression model tests were run with each of those variables that had a univariate association with ED visit or hospital readmission at a p<0.20 level to test the predictive value of each target variable for (1), (2), and (3) above. In all three logistic regression models, the independent variable of interest—that is, type of referral (e-485 or handwritten)—was significant as a predictor variable in relation to whether or not there was an ED visit:

  1. Emergency department visit (R-square = 0.057)—Variables included: age (B= −0.178, p=0.776); cardiovascular accident (B= −18.643, p=0.998); referral type (B= −1.278, p=0.012).
  2. Hospital readmission (R-square = 0.051)—Variables included: age (B= −1.224, p=0.016); referral type (B= −0.843, p=0.069).
  3. Either emergency department visit or hospital readmission (R-square = 0.049)—Variables included: age (B = −0.879, p = 0.076); referral type (B = −0.983, p = 0.020).

As shown above, the findings from all three regression analyses show that the type of referral was significant in the prediction of whether or not subjects were seen in the emergency department or readmitted to the hospital within the 30-day period after referral to home health care:

  • For (1) above, referral type was, in fact, the only significant predictor; age, gender, and the various comorbidities were not significant.
  • For (2) and (3), referral type and age were predictor variables.

It is clear that the use of e-485 is associated with a lower likelihood of readmission to the hospital or visit to the emergency department. That is to say, when subjects were referred with the e-485 rather than the handwritten form, it was more likely that they would not be readmitted to the hospital or visit the emergency department within the first 30 days after referral.

These results were true in both the unadjusted and the adjusted analysis. The unadjusted odds ratio for ED visit was 0.26 (p<0.001), and the adjusted odds ratio was 0.28 (p=0.005). Likewise, the unadjusted and adjusted odds ratios for hospitalization were 0.36 and 0.43, respectively (p=0.07).

To explain the relationship between readmissions/ED visits and referral method, the study team hypothesized that the improvements in the quantity and comprehensiveness of process measures led to better care. Significant improvements in orders for medication, allergies, functional impairment, physician notification, goals, and nursing orders would logically have an impact on treatment. This hypothesis is the central to initiatives that use technological tools to effect positive change in health care. Given the data presented above, it may be worthwhile for future studies to pursue the effects of such an intervention as the e-485 on a larger scale and perhaps in conjunction with other such technological tools.

Pre-post intervention study. The control group for this study consisted of patients referred from Weill Cornell-NYPH to VNSNY in the 12 months prior to the initiation of the e-485 from the sites in which the e-485 was available at Weill Cornell-NYPH. The intervention group consisted of patients referred to VNSNY during the subsequent 19 months using the e-485.

The study team had initially intended to examine the completeness of key information on home health admission and compare home health outcomes of the pre- and post-intervention samples using routinely collected data from the Outcome and Assessment Information Set (OASIS). However, after initial analyses had shown little difference in the completeness of patient data on admission, investigators found that CAU nurses work with referring physicians and their staffs to complete missing orders and to verify diagnoses and insurance information for any patient that qualifies for home care. This completion of the information transfer is done by phone, and the final order set is significantly different from the initial referral order set.b

Therefore, because the differences between the pre- and post-samples in the completeness of most data on admission were small, the scope of the examination of OASIS outcomes was reduced. Any differences observed could not be attributed to the introduction of the e-485, so the comparisons were focused on the completeness of the disease-management orders built into the e-485 as well as on whether the medication prompt in the e-485 lead to higher rates of evidence-based pharmacotherapy in the post-intervention sample.

The sample size for the post-intervention sample in the pre-post analyses is slightly lower than in the previous analyses because of patients' ineligibility for home care. The sample size is further reduced in analyses of evidenced-based orders because a small number of the e-485 forms in the post-intervention sample were sent to CAU for a resumption of care following an inpatient stay and not for an initial admission. These forms were taken out of the analysis.

Items from the Outcome and Assessment Information Set were linked to patient referral data to measure patient clinical status on admission. Evidence-based disease management and medication orders were obtained from electronic records maintained by VNSNY. At least one disease-specific, evidence-based order had to be on the initial plan of care for evidence-based orders to be classified as complete. (Go to Appendix A for descriptions of specific disease management and evidence-based orders.)

Table 4 compares the demographic and diagnostic characteristics of the pre- and post-intervention samples and Table 5 presents results of the pre-post analysis. 

Table 4. Characteristics of pre- and post-intervention samples

CharacteristicPercent in pre-intervention sample (n)Percent in post-intervention sample (n)
n294122
Female70 (205)70 (85)
Age:  
Under 6532 (95)13 (16)
65 < 7512 (35)6 (7)
75 < 8529 (86)20 (24)
85 and over27 (78)61 (75)
Diagnoses:  
HIV/AIDS19 (57)12 (15)
Congestive heart failure19 (28)12 (14)
Diabetes22 (64)17 (21)
Wound5 (14)9 (11)
Hypertension33 (98)35 (43)
Diabetes with hypertension50 (32)9 (11)

Table 5. Results of the pre- and post-intervention analysis

 Pre-samplePost-sampleP
Completeness of evidence-based orders:   
Congestive heart failure (up to 5 diagnoses)  11% (n=28)27% (n=11)0.324
Diabetes (up to 5 diagnoses)86% (n=64)81% (n=21)0.727
Medication order1 for congestive heart failure patients11% (n=28)45% (n=11)0.028
Medication order1 for diabetes with hypertension diagnosis34% (n=32)45% (n=11)0.719
Physician involvement in generating orders and billing3% (n=274)88% (n-109)<0.001
Average length of home health stay, in days76 (n=290)82 (n=119)0.630
Average number of skilled nursing and therapy visits24 (n=290)29 (n=119)0.228

1 Medication order is defined as beta-blocker at time of admission for congestive heart failure (up to 5 diagnoses) and ACE inhibitor at time of admission for diabetes with hypertension (up to 5 diagnoses).

It is noteworthy that physician involvement in generating home health care orders and billing for home care was significantly greater in the post-intervention sample.

As part of the evaluation, feedback was gathered from physicians at Weill Cornell-NYPH as well as the intake coordinators at the VNSNY. Specifically, Weill Cornell-NYPH physicians reported that the e-485:

  • Required an average of 3 minutes to complete.
  • Was accessible as part of the medical record.
  • Was easy to learn.
  • Enabled the physician to bill for managing home health care patients.

Users at Visiting Nurse Service of New York reported the following impressions of the e-485:

  • Form was easy to read.
  • Format was easy to follow.
  • Orders were more comprehensive.
  • Titration orders facilitated co-management of patients by nurses and physicians.

Investigators concluded that use of the e-485 for referring patients to home care was associated with higher physician involvement in generating orders for home health care and a higher probability of being prescribed a beta-blocker for patients with congestive heart failure. However, differences in the use of home care services (either in the number of home health visits or in the average length of home health service days) could not be found. The higher use of beta-blockers is consistent with the e-485 having a built-in alert to physicians to prescribe a beta-blocker for patients with a known diagnosis of congestive heart failure.


aThe printed CMS Form 485 itself is no longer required but all elements from the form must be included in a home health care plan. Go to Appendix A for a list of elements in the traditional Form 485.
b This "work around" was developed in an environment where initial referrals often are very sketchy. The e-485 presents an opportunity to greatly improve the efficiency of the referral process. It may, however, have less impact on the completeness of at least some types of referral information at the time the patient's nurse makes the first visit.

Current as of September 2007
Internet Citation: Development of Electronic Transition Tools for Home Health Care: Final Contract Report. September 2007. Agency for Healthcare Research and Quality, Rockville, MD. http://www.ahrq.gov/professionals/systems/long-term-care/resources/hcbs/etransitions/etransitions.html