Improving Patient-centered Care Using an Inventory

Research Design and Methods: This study will utilize a multi-phase mixed-methods approach to develop and test a patient inventory to facilitate patient-centered care delivery. In Phase I the investigators will identify which factors should be considered for inclusion in the patient inventory. This will involve interviews and focus groups with providers and patients, respectively; and a review of the relevant literature. In Phase II the investigators will develop and test an inventory using a Delphi panel and assembling a questionnaire repository. In Phase III the investigators will conduct a randomized trial comparing the patient inventory to a control condition (e.g., healthy living video) to assess whether use of the inventory results in care that is more patient-centered and decisions that are more collaborative. Phase I: To identify patient contextual factors and preferences, and patients' experiences, awareness, resources, and willingness to use technology for care delivery, the investigators will conduct focus groups with patients at 2 facilities and 4 CBOCs in Veterans Integrated Service Network 12 (VISN 12) representing urban and rural settings and a mix of sociodemographic characteristics. In order to capture the perspective of Veterans who live in more rural areas and who may represent a broader range of socio-demographic characteristics, the investigators will conduct focus groups at four CBOCs associated with the medical facilities that represent urban/rural and ethnic diversity. The investigators plan to recruit both patients who have received care through technology (e.g., home telehealth (HT), CVT) and those who have had little or no experience with care delivered through technology. The investigators will conduct separate focus groups with patients who have received care through technology and those who have not used technology for health care. Focus group discussions will be transcribed verbatim and analyzed using constant comparative techniques to code the data thematically. This information will serve as the foundation for developing the inventory. A total of 8-12 focus groups will be conducted with 5-8 Veteran participants per group (n=40 to 96 patients). The investigators will conduct 2 focus groups at each VA facility, and 1-2 focus groups at each CBOC. Based on prior experience conducting focus groups with Veterans, the investigators believe that this number of focus groups will provide sufficient data to reach saturation, the point at which no new information is uncovered during group discussion. Two members of the research team will manage each focus group. Patient Focus Group Questions. A discussion guide that includes open-ended items with probes and follow-up questions will be used to conduct focus groups with patients. The guide consists of questions designed to generate discussion among participants concerning contextual factors, patient-centered care, and technology use for health care delivery. The first set of questions will assist in identifying contextual factors that affect patients' abilities to manage their care. The second set of questions will explore patients' understanding of and preferences for patient-centered care. The final set of questions will describe patients' awareness of, available resources, and willingness to use health technology. Data will be analyzed using constant comparative techniques consistent with a grounded-theory approach Provider Interviews. Semi-structured interviews will be conducted with healthcare providers at each of the sites. Providers will consist of PACT and/or primary care team members including nurse care managers and physicians. The investigators will use semi-structured interviews for providers because of the added flexibility afforded by this method with respect to time and scheduling. The investigators developed a semi-structured interview guide to address the research objectives. The first set of questions is designed to identify contextual factors and understand how providers attempt to address them. Data will be analyzed using the same process detailed for analysis of focus group data (i.e., constant comparative techniques consistent with grounded theory). Question Repository. In parallel with focus groups and interviews, the investigators' research assistant will review the available literature through PubMed, psycINFO and other reference search engines for surveys/questionnaires that include items related to areas of patient context. Initial search terms include: assessment tool, assessment scale, risk assessment, contextual factors, patient questionnaires and context specific terms such as technology resources, social support and patient-centered care. The investigators will assemble a repository of questions that could be used for the investigators' inventory. If there are contextual areas in which the investigators are unable to find adequate questions, the investigators will generate items for consideration by the Delphi panel. Phase II: Creating the Patient Inventory. Using findings from focus groups, information from the literature, and input from an expert panel the investigators will draft a patient inventory that can serve as a collaborative decision tool for patients and providers. The investigators will work with the expert panel to identify and refine the content areas for the inventory using a modified Delphi panel approach. The goal of the project is to develop an instrument that would take no more than 10-15 minutes for a patient to complete. Participants/procedures. A heterogeneous panel will be chosen based on participants' willingness to participate and knowledge of the relevant topic. Because patients are the key stakeholder for patient-centered care, the investigators will assemble a panel that includes Veteran representatives, VA providers, researchers and representatives from the investigators' operations partners. To maintain at least 14 respondents (70%), the investigators will initially recruit 20 panelists (5 in each stakeholder group). The investigators will conduct the Delphi process through a series of email exchanges. In the first email round, a list of possible items for each of the context areas will be distributed. Panelists will be asked to rank the importance of the areas of context identified, comment on whether they feel that there is redundancy in categories, and provide input on the relevance of items within each content area to patient context, patient-centered care and/or collaborative decision-making. The investigators will also ask an open-ended question allowing panel members to identify areas that they feel should be included. Responses will be entered without identification of the panelists. Each participant will be assigned an ID code. Categories and items for which the panelists agree do not fit or are redundant with other items that are better worded, and those items that are ranked very low will be dropped. A revised set of categories and items will be sent to the panelists for a second round of review. At that time, the panelists will repeat the process used for the first round review. A final draft will be sent to the panel for one more review. The investigators will also include feedback about qualitative comments made by panelists in each round and data about specific item ratings (e.g., median, lowest and highest ratings per item, etc.) The responses to this last review will be used to finalize the patient inventory. The investigators anticipate that this process will take approximately 4-5 months to complete the Inventory Template. Once the patient inventory is finalized, the investigators will integrate them into an iPad application for patient use. The iPad provides multiple advantages for collecting patient information, providing a portable, touch-driven multimedia experience popular across different demographic strata. The iPad inventory will include skip patterns for respondents to jump over non-relevant items, out-of-range checking and validation so that respondents do not enter erroneous information, drop-down boxes, and utilize examples and explanations for clarity. Patients will have the opportunity to "not answer" any items, based on personal comfort of what information should be shared. The investigators will perform iterative testing of the iPad version of the inventory to ensure acceptability by patients. All information entered in the iPad will be kept private and secure, such that other patients cannot access any stored data. As an additional safeguard, the iPad will "time out" if not used for some duration of time to prevent extraneous viewing by others. The iPad will be used in waiting areas by patients prior to encounters with health care providers. Research staff will maintain the iPad, including charging, storing securely when not in use to prevent theft, applying anti-bacterial cleaner, and monitoring device use. When patients use the iPad inventory, the application will create a brief report summarizing patient preferences and contextual factors. This report will be available to the PACT at the point of care to facilitate patient/provider communication and collaborative decision-making. For the initial study, research staff will facilitate communication of the report to health care providers by generating a paper summary that patients will hand to providers. However, design and testing will consider future implementation strategies, including incorporation into clinical workflow patterns and adoption by clinical staff. Phase III: Testing the Patient Inventory. The investigators plan to pilot the patient inventory in a sample of Veterans in two sets of PACT clinics; one at Hines and the other at JBVA. The study design is a randomized controlled trial. The investigators will randomize half the patients to complete the patient inventory and the other half to a control condition (i.e., health video). The investigators will stratify by Veterans Affairs Medical Center (VAMC) to ensure that there are equal samples. The randomization scheme will be developed using computer generated random assignment. Eligible patients will be identified and recruited from the PACT clinics at Hines and Jesse Brown VAMCs. No subjects will be excluded based on gender, race, or ethnicity. Eligibility criteria include: (1) adult, age 18 or over; (2) assigned primary care provider in PACT clinic; (3) no history of dementia; (4) no blindness (unable to view iPad app). Only patients of participating providers will be eligible for the study, to ensure both patient and provider consent to audio-recordings and study procedures. It is possible that providers will discern that their visits with patients that involve the use of 'summary' forms are the same visits that will be audio-recorded. To address the likelihood of a Hawthorne effect the following strategies will been undertaken. First, providers will be informed that some of the visits that are recorded will include patients providing additional written information for providers while other visits will not include this information. Second, the investigators will undertake sensitivity analyses comparing data from the first 10% of audio-recorded encounters to the 2nd and 5th 10% of encounters. The investigators will assess whether provider behavior is significantly different at these points in time. If the first several visits result in different behavior then later visits, the investigators can eliminate the first recordings from their analyses. Thus, the investigators will over-recruit patients by 10%, with patients evenly distributed between intervention and control arms. Intervention: The patient inventory developed through focus groups and Delphi panel will serve as the intervention. The inventory will be deployed using an iPad. Patients will provide a copy of the summary to their provider at the beginning of their visit. Control Condition. To address concerns about a Hawthorne effect, the investigators will implement an attention control. The control group will use the iPad to will view a brief health video (e.g., 8-10 minutes) containing general health information. The VA National Center for Health Promotion and Disease Prevention (NCP) is currently developing a set of video-based health living messages. The investigators will utilize one of these videos in for their control group. Alternatively, Agency for Healthcare Research and Quality (AHRQ) has developed several videos that are available at no cost and are 508 compliant. The 'Be involved in your health care - Questions are the answer' is a 7 minute video that is focused on talking to your provider and asking questions (http://www.ahr.gov/questions/video/waitroom/). Control patients will also be asked to audio-record their encounter with their provider, but not to reveal that they are doing so. Patients will complete the informed consent process, agreeing to use of the iPad to either complete an inventory or view a video about their health, willingness to have their encounter with the provider audio recorded, and completion of a set of questionnaires following the clinic visit. Consenting patients will complete the patient inventory or view the video on an iPad prior to their clinic visit. A research assistant will instruct the patient on how to use the iPad and what the patient needs to do to complete the questionnaire (in the intervention condition). The RA will be available to answer questions or provide additional instruction, as needed. After intervention patients complete the questionnaire, the RA will hook the iPad up to a portable printer so that a hard copy of the inventory can be generated for patients to immediately give their provider. The patient will proceed to his/her visit with the provider. To ensure adequate use of the iPad and the patient inventory, the RA will be available to assist subjects in use as necessary. The research team will collect information on how frequently assistance is needed, and form of support provided. The investigators estimate the inventory will take approximately 10-15 minutes to complete. Patients will have the opportunity to complete the inventory in a designated, private area to reduce interruptions and distractions. In both the intervention and control arms, patients will be asked to conceal a small recording device in their pocket and/or bag in order to record their conversation with the provider. The patient will be told that the provider has agreed to be recorded during some of their patient visits but that the provider will not know which visits will be recorded. The RA will turn the device on before handing it to the patient and the patient will return the recorder after the visit. Weiner has successfully utilized this strategy in two HSR\&D funded studies. Following the clinic visit, the patient will be asked to complete a set of questionnaires. These items will be completed using paper and pencil. The investigators will also ask patients to indicate whether they have used any technology for their health care using a checklist (use of home telehealth, CVT, secure messaging, prescription refills through HealtheVet \[MHV\], use of mobile apps like PTSD coach, accessing VA Facebook or twitter sites) and whether any future care plans include use of technology. Providers. PACT providers will also serve as research subjects for the study. The investigators will ask providers to consent to having some of their encounters recorded over a one year period. The investigators will explain to providers that they are interested in examining the effects of providing or not providing additional patient-generated information to providers at the time of their visit on patient/provider interactions. Providers will be assured that no identifiable information will be recorded on the transcriptions (e.g., providers will be referred to by a unique study ID number). The investigators will also ask providers to participate in a debriefing interview at the end of the study to ask them about the value and use of the pre-visit patient inventory in their encounters with patients. Only patients of providers who consent to participate will be approached about study participation. Study Measures: To determine patient-centeredness the investigators will ask patients to tell them about their experiences with care and the extent to which care they received met their preferences and addressed their contextual factors. The investigators will utilize measures of patient-centeredness including the Consultation and Relational Empathy (CARE) measure and the Consultation Care Measure (CCM). To objectively evaluate the extent to which care being provided is patient-centered, the investigators will obtain audiotapes of patient/provider encounters and code them using Weiner's 4C analysis measure and the Informed Decision Making (IDM) tool to evaluate collaborative decision making during patient/provider encounters. These measures are described below: CARE. The Consultation and Relational Empathy (CARE) Measure is a process measure of empathy and holistic care in the context of a therapeutic relationship. Mercer defines empathy in the medical encounter as the ability of the provider to understand the patient's situation, perspective and feelings; to communicate to the patient what their understanding is and check whether s/he is correct in this interpretation, and then to act in a way that is helpful and therapeutic .This description fits into many definitions of patient-centered care. The CARE measure has 10 items and it specifically asks about today's consultation. To better fit the definition of care in the US, the investigators will change the word consultation to 'visit.' For each item, the respondent rates their response as poor, fair, good, very good, excellent, or does not apply. Questions include: How was the provider at 'making you feel at ease', 'letting you tell your "story"', 'being positive.' The score for each item is added for a maximum score of 50 and minimum of 10; a higher score reflects greater 'patient-centeredness'. CCM. The Consultation Care Measure (CCM) measures a patient's perceptions of patient-centered care during their last visit with their provider. There are 5 subscales in this 21-item scale. CCM assesses communication and partnership (11 items), personal relationship (3 items), health promotion (2 items), positive and clear approach to the problem (3 items), interest in effect on life (2 items). The items use a 4-point Likert scale for scoring from 1= very strongly agree to 2= strongly agree to 3=agree to 4=neutral/disagree. Cronbach's reliabilities range from 0.84 to 0.96 for the subscales. A higher level of patient-centeredness based on the CCM (lower score) was associated with short-term better health outcomes. Informed Decision Making (IDM) measures the extent to which decision making involving the patient and provider includes 7 elements of a complete decision. The 7 elements include: discussion of patient's role in decision making, clinic issue or nature of the decision, alternatives, pro and cons (benefits and risks), uncertainties, patient understanding, and patient preference. For each decision, a score between 0-7 is assigned based on review of the audio-recorded encounter. A higher score is expected if the decision is more complex (decisions can be categorized as basic, intermediate or complex). For basic decisions (e.g., refilling a medication), a score of 2-3 would indicate a good IDM event, while a more intermediate decision (e.g., prescribing use of a new piece of equipment - such as CPAP) should result in a score of 5-6 reflecting an informed decision; and a complex decision (e.g., lifestyle counseling) would receive a score of 7 if all the elements of the IDM were noted during the decision. Content Coding for Contextualizing Care (4C). The 4C analysis is methodology for assessing patient centered clinical decision making. Specifically, 4C rates encounters based on whether the provider (a) probes for clinically relevant context and (b) plans the care accordingly. The analysis is generated from both the patient's medical record and from an audio recording of the provider-patient visit. 4C may be conducted by as few as two individuals or as many as five, with the additional personnel required to verify inter-rater reliability and other sources of evidence for validity. Coding will be done by an experienced coder who has worked with Dr. Weiner on two prior studies utilizing the 4C analysis. Using this coding we will determine how often providers probe for contextual factors during patient encounters, and how often they incorporate this information into their care planning for that encounter. Sample size/power: The primary measure of patient-centered care is the CARE measure. Normative data on the CARE indicates a mean score of 40.8 (range 10-50) with a between-physician variance of patients receiving care from the same physician, = 9.0247, and within-patient variance = 69.2149, for a total variance = 78.2396 (SD=8.85) (Mercer 2005). Without taking into account related patient CARE scores of patients being treated by the same physician, 64 patients in both control and intervention groups (total n=128) will be adequate to detect a moderate effect size (Cohen's d=0.5) at the 0.05 level with 0.8 power. Of course, patients receiving care from the same physician might respond similarly to questions regarding empathy and holistic care pertaining to their physician, as recorded by the CARE measure. The multiplicative effect on the standard error due to clustering is 1 + (m-1)\*ICC, where m represents the size of each cluster (patients per physician) and ICC the intraclass correlation of same-physician patient CARE outcomes. 25 doctors with 11 patients each, or 24 doctors with 12 patients each will be sufficient in detecting a moderate effect size (Cohen's d=0.5) at the 0.05 level with 0.8 power. Accounting for the possibility that patients may change their mind, be too fatigued, and/or have not enough time to complete the post-visit questions, or deficiencies in physician attrition, the investigators have added an additional 10% (28 more patients/25-26 physicians) to the total. The total study sample size is 25-26 providers and 300-308; 150-154 patients per group. Patients enrolled in the PACT clinics at Hines and Jesse Brown will serve as the patient subject pool and the providers in these PACT teams will serve as the provider subjects. Only patients who are in the panels of providers who consent to participate will be recruited for the study. Analysis: Descriptive statistics including frequencies and means will be calculated to characterize the investigators' patient sample, including demographics (e.g., mean age, gender, ethnicity/race). The investigators will utilize t-tests to make comparisons of intervention and comparison groups on the patient measures. The overall scores and the subscale scores for the CARE and the CCM will be compared using t-tests. They will then examine the extent to which other characteristics including study site (Hines, JBVA) and patient characteristics (age, race) account for differences in ratings of patient-centered care using linear regression models. The investigators will use the standard formula Y = a + b1X + b2X +. biX, with no more than 15 covariates (using 1 covariate for every 10 observations) in addition to the intervention effect (b1). As many patients may have the same provider, there could be an effect for provider. They will explore use of multi-level modeling to account for provider effects. 4C Coding and Analysis. 4C analysis rates encounters based on whether the provider probes for clinically relevant context and plans care accordingly. The analysis uses both the patient's medical record and an audio- recording of the provider-patient visit. 4C coding includes the following: 1. Contextual red flag identification: Following an encounter, the first step for coders is to determine if there were any contextual red flags that the provider would need to have explored to provide appropriate care. Red flags include things like missed appointments, non-adherence of medications or receipt of laboratory tests or deterioration of chronic illness (e.g., uncontrolled blood pressure). A contextual red flag is defined as anything a patient says or that was observed about his situation or behavior that suggests that contextual factors were not addressed and may be contributing to problems with their care. The search for contextual red flags begins with a structured medical record review by a contextual red flag "screener". If no flags are found in the medical record the coder listens to the audio-recording to determine if the patient made statements suggesting underlying contextual issues essential to their care. For example: a. Contextual red flag identified from chart review: Rising systolic blood pressure in patient with hypertension who has been previously well controlled. b. Contextual red flag identified from audio-recording: Visually impaired Veteran with hypertension mentions that his daughter, who sets up his medications in a daily pill box, has relocated to another city for her job. 2. Coder(s) formulate an unambiguous direct probe of red flag: The coder(s) compose an unambiguous provider probe to the identified contextual red flag, e.g. "I notice that your hypertension used to be well controlled but your blood pressure has been very high over the last couple of months. What do you think is going on in your life that might be a factor in this problem?" The probe is always in the form of a question by the provider in response to a red flag. The purpose of this exercise is to help the coder identify what s/he should be listening for to indicate that the provider has noticed and is pursuing the contextual red flag. 3. Coder listens for clinician probe: The coder then listens for whether the provider pursued the contextual red flag in a manner that substantively approximates their direct probe. 4. Coder listens for a 'patient reveal' in response to a probe: If the provider probed, the coder determines whether the patient revealed contextual information that is relevant to care, such as "It's a lot more difficult for me to remember to take my medication when I'm supposed to since my daughter left." 5. Coder formulates an unambiguous direct response to the patient reveal: If the patient reveals contextual issues in response to the probe, the coder formulates an unambiguous response to the reveal that indicates the provider recognized the need to contextualize the care plan, e.g. "Let's talk about how you could change your medication schedule to help you remember." 6. Coder listens for contextualization of care plan: The coder then listens for whether the provider recommends a plan of care that substantively approximates their direct probe. (e.g., switches all medications to morning) Because coding can be quite nuanced, Weiner et al. has developed a comprehensive training manual, similar to those developed for other coding systems, such as Debra Roter's "Roter Interaction Analysis System" (RIAS), utilized for coding physician patient communication. RIAS codes interactional process behaviors (e.g. when clinician or patient asks open vs closed ended questions, gives instructions, asks for understanding, paraphrases, etc.) but not content. The 4C coding manual is available from Dr. Weiner, co-inv on the study. (http://dvn.iq.harvard.edu/dvn/dv/4C) IDM Analysis. Reliability of coding between three coders will be established with at least 85% concurrence before the coders will individually code encounters. The investigators will recheck reliability after every 25 encounters by having all coders code the same encounter to ensure that they are all coding in a similar manner. If concurrence is below 75%, the investigators will review coding criteria and re-establish concurrence using the next 2-3 audiotaped encounters. Coders will listen to the audiotapes and identify the number and types of decisions that occurred during each encounter. Then they will select the decision of more interest/concern of the patient and code that decision using the IDM criteria. The investigators will compare the mean IDM scores by decision type by intervention vs. control groups to determine whether use of the inventory facilitates more informed decision making Cost impact: In this study, the investigators will conduct a cost-identification analysis to assess the cost of providing the intervention in this study. Additionally, they will conduct a budget impact analysis to assess the impact on costs to a VA medical center of implementing the patient inventory. The cost of the intervention includes the cost of the equipment needed for the intervention (i.e., the iPad and printer) and the cost of any change in clinic personnel time that results from the intervention. The patient inventories may impact clinic personnel's time due to the need to train and assist the patient in entering the inventory as well as potentially changing the duration of the clinic visit due to discussions facilitated by the patient inventory. The RA will record 1) the time it takes to train each patient to use the iPad, 2) the time it takes for each patient to complete the patient inventory and the proportion of this time the RA spent assisting the patient, and 3) the time it takes to print the inventory for the patient to give to the provider. Additionally, the RA will record whether it interferes with being on time for the clinic visit. The RA will also record the length of the provider/patient encounter by recording the time the provider entered the clinic room to see the patient and then when he leaves the room/completes the visit. In order to make comparisons to usual care, the RA also will record the length of time of usual encounters by timing the encounters of the control patients in the randomized. The investigators will also conduct a budget impact analysis to evaluate the cost of implementing the intervention at a VA medical center. The estimated budget impact on the medical center will be sum of the purchase price of the equipment, the cost of the clerical personnel's time, and the cost of physicians' time (which will be estimated as an average physician's wage multiplied by the difference in the average time of clinic visits between intervention and control patients). The investigators will estimate the total cost of providing the intervention to patients under the alternative scenarios regarding the number of patients who would complete the inventory and the amount of equipment that would need to be purchased.