The setting
The study was done in 3 dialysis units managed by The Rogosin Institute (New York, NY), affiliated with New York Presbyterian Hospital and Weill Medical College of Cornell University. All provide treatment by in-center chronic HD. Unit A also trains and treats patients by peritoneal dialysis, home HD, and nocturnal self-HD. Eight full-time salaried Rogosin Institute staff nephrologists, who also care for renal and transplant patients, teach, and do clinical research, and 2 nurse practitioners care for patients in Unit A. In Unit B, there are 1 to 2 Rogosin nephrologists, 2 to 5 in private practice, and 1 nurse practitioner. In Unit C, the 3 nephrologists are in private practice.
Participants
The patients were all 1790 patients treated by chronic maintenance HD from 91 days after ESRD start.
Design of the study
The electronic patient record (Disease Manager Plus™, MIQS® Inc, Boulder, CO) employs a relational database (Sybase® Adaptive Server Enterprise, Sybase, Inc., Dublin, CA), running on a server computer (Sun® Microsystems, Santa Clara, CA). It is accessed using a custom toolset (4D, Inc., San Jose, CA) from client personal computers in dialysis units, renal and transplantation practice, physician offices, hospital, and home. All clinical, administrative, and financial information is immediately accessible at all times on patients ever entered into the database. It serves all kidney disease care, including dialysis and transplantation. Subject to security considerations, lifetime patient data relevant to pertinent caregiver needs are accessed whenever and wherever needed. The security ensures confidentiality for clinical and financial information and for integrated electronic mail. Laboratory test results, radiology reports, pathology reports, and dialysis machine data download automatically into the database via MIQS-designed electronic interfaces. Dialysis machine data enable on-line chair-side and remote real-time monitoring including home nocturnal HD [21].
Coded data elements include diagnoses, procedures, symptoms, signs, medications, allergies, and hospitalizations. Patient-specific ICD-9-CM codes record reason(s) medications are prescribed and patients admitted to hospital. Notes are charted in free text or using templates. Advance directives, living will, do not resuscitate, treatment consent, and other documents are stored in the database and readily accessible. HD treatment screens record all details. HD orders and medications to be given during HD automatically populate treatment screen fields.
To provide clinically useful point-of-care reports embedded query tools are incorporated to organize data quickly in any way desired over any time period, to make knowledge available about individual patients, and groups. Reports that can be updated and organized at the point-of-care are user-designed to facilitate clinical decisions based on timely, complete, relevant, patient-specific, time-oriented data.
The electronic starting point for patient encounters displays all relevant historical information including reports, medications, allergies, and patient-specific and rules based alerts and reminders. Encounters, tailored to specific functions such as HD, peritoneal dialysis, chronic kidney disease, and transplantation facilitate data entry and communication with others, e.g., referring physicians. Individual patient reports accessible on encounter and HD treatment screens include contemporaneous medications, comprehensive lifetime lists of diagnoses, surgical procedures, diagnostic procedures, allergies, adverse drug reactions, immunizations, and hospitalizations (Figures 1 and 2). Others display data over time in spreadsheet format from domains including signs, symptoms, medications, laboratory tests, HD orders and treatments, diagnoses, procedures, and hospitalizations (Figure 3).
Patient treatment groups, a special software functionality, define patients receiving treatment courses by in-center HD, other dialysis modalities, kidney and pancreas transplants. An in-center HD group is defined by dates of first and last HD in a treatment course. Reason the course ended is charted from a coded list that includes patient expired, recovered renal function, transferred to another dialysis unit, transferred to another group, e.g., kidney transplant.
Data Analysis
Reports that incorporate information from multiple information domains were used for data analysis on patient groups. An integrated patient selection query tool enabled selection of cohorts for inclusion in reports. Among selection criteria were demographic elements, locations, alive or expired within a defined time range, presence or absence of ≤3 patient groups and ≤3 ICD coded diseases and procedures.
The Treatment History with Adjusted Dates report enabled much analysis. It adjusted dates automatically to first and last days of the chosen period, calculated days in the group, and displayed why the group ended. For the present analysis it was modified by adding age (at treatment history start), ethnicity, gender, ESRD date (first ever treatment by dialysis or transplantation), date of death, primary cause of ESRD. ESRD vintage was calculated as (Start date of dialysis in the period of study – ESRD date).
The report was generated using the patient selection query tool to select the pertinent HD cohort, Unit A, B, or C, individual calendar year or full 6–9 year spans. Days in the group were summed and average age calculated. Saved as an ASCII file, it was imported for further analysis into Microsoft® Excel® or SAS® JMP™.
USRDS reports data only from the 91st day after start of the first dialysis treatment [12]. To enable comparable analyses, data were sorted first by (Start date – ESRD date), and second by treatment group days. Patients starting dialysis within <91 days and with <91 dialysis treatment days were excluded. Those starting dialysis within 91 days of ESRD date were included if treated for >90 days; treatment group days were reduced by days prior to the 91st, as in USRDS.
The reason treatment was discontinued was examined, using the following conventions:
• Kidney transplants: Patients were deemed alive at the end of the prior dialysis course.
• Deaths: Date treatment was discontinued because patient expired was checked with date of death recorded under patient demographics. Known death within 30 days of transfer to hospital, nursing home, hospice or another dialysis unit was ascribed to the previous modality.
• Transfers to other units for continuing dialysis care: The database was searched for site of future dialysis care, and latest recorded patient-caregiver contact. Time from transfer to last documented contact was calculated. Patients documented alive ≥ 6 months after transfer were treated for analysis as alive at relevant study year-end. Patients with no such information available were considered "lost to follow up"
Calculation of Mortality
Mortality per 1000 patient years was calculated for each individual study unit in each calendar year, as is done by USRDS, by dividing the number of deaths in each individual year by the total time in years that the patients were treated by HD in that calendar year, and multiplying by 1000.
Data Verification
Analyses were made on several occasions. Dubious or missing values were checked and corrected as necessary in the EPR. Many coded values were missing because data had been charted as text only; coded values were entered from physician text notes. For final analysis, reports were run by individual calendar years and the entire 7–9 year period.
Study Approval
The study was approved by the Institutional Review Board of Weill Medical College of Cornell University.