There is a total of 2364 data entries which constitutes 1602 patients. Of the total patients, 462 are patients who have been admitted more than once. This is displayed in Figure 4.2. A total of 22 entries were deleted from the original dataset owing to them being incomplete or duplicate entries.
Readmission is generally measured as occurring within either 30 or 90 days. Readmission within 90 days is a global measure for the quality of service given after discharge, but it is arbitrary and one cannot be sure how many admissions a patient may have had for example, at another level 1 hospital before being admitted to Stikland Hospital (Koen & Smit, 2016a). The clinical
4.2. The Stikland dataset 77
Table 4.2: Variables evaluated and included in the analysis.
Variable Original admission dataset Original disc harge dataset Re-captured Additionally captured Deriv e d Inadequate information Included in final datase t Variable Original admission w orkb o ok Original disc harge w orkb o ok Re-captured Additionally c aptured Deriv ed Inadequate inf o rmation Included in final da taset
Admission date[LOS] [Daysdischarged]
1 1 Secondary ICD-10 de-
scription
1 1 1
Folder number 1 1 1 Secondary ICD-10
code
1 1
Area admitted from* 1 1 1 Crisis-discharge 1 1
Ward 1 1 Transferred to 1 1
Voluntary admission 1 1 1 Date of birth[Age] 1
Assisted admission 1 1 1 Age* 1 1
Involuntary admission 1 1 1 Follow up* 1 1
State patient 1 1 New Beginnings
and/or ACT*
1 1
Day patient 1 1 1 Readmission (Y/N)* 1 1
72 hour observation 1 1 1 Total readmissions for
specific year
1
First admission 1 1 Total admissions ’12-
’14*
1 1
Readmission 1 1 Readmission count* 1 1
Readmission in 90
days
1 1 Days discharged be-
fore readmission*
1 1
Income level 1 1 Readmission within 30
days* 1 1 Re-admission after conditional discharge 1 1 Readmission within 90 days* 1 1
Transferred from 1 1 Length of stay* 1 1
Discharge date [LOS] [Daysdischarged]
1 ICD-10 description* 1 1 1 1
Primary ICD-10
description[Diagnosis]
1 1 Substance abuse* 1 1
Primary ICD-10 code 1 1
[...]
Derived from the variable *Used in analysis
SMEs defined a revolving door patient as one having more than one readmission, and suggested a categorical variable for a patient having either none, one or more than one readmission. Readmission within 30 and 90 day will also be evaluated. A patient’s first entry in the dataset will be regarded as the patient’s first admission assuming they have never been admitted before, even though this might not be the case. This decision is made owing to not having the previous years’ data for the patients. The dependent variable for the data analysis will be whether a patient is readmitted (1) or not (0). The readmission criteria as defined by the clinical SMEs (0,1 and > 1) along with readmission within 30 days (1/0) and readmission within 90 days (1/0)
Figure 4.2: Total number of admissions of a patient during the study period.
will also be investigated. The independent variables are described as being the following 1 :
Admitted from [Area] This categorical variable was reduced from 78 recorded areas to five classes by the psychiatrist and admission nurse:
• [1] Paarl, Vredenburg and surrounds; • [2] Eerste Rivier Hospital and service-area; • [3] Karl Bremer Hospital and service-area; • [4] Direct admissions (Stikland); and
• [5] Other areas of which the majority are not in Stikland’s jurisdiction Follow-up: The place of follow-up was captured as a mixture of codes and descriptions. In
the case where a code was unknown, which was the case with only three codes, a sample was drawn from the archive by the clinical SMEs to determine what the code represents. Many codes and descriptions were found to be for the same use. Afterwards there were 19 follow-up possibilities which were further grouped into eight categorical variables namely:
• [PHC] Primary healthcare clinic;
• [Stikland] Stikland Psychiatric Hospital; • [Tygerberg] Transferred to Tygerberg;
• [Other] Other (long term wards, other provinces, private care, passed away); • [None] No follow-up;
• [NB] New Beginnings institution; • [ACT] ACT programme; and
• [ NB&ACT] New Beginnings and ACT.
1
Dependent variable title[Variable name in dataset if different than title] description [class name in dataset] description
4.2. The Stikland dataset 79
Community programme: [Instit] Another categorical variable describing whether a patient was admitted to any community programme or institution:
• [NB] New Beginnings institution; • [ACT] ACT programme;
• [NB&ACT] New Beginnings and ACT; and • [None] Do not belong to any programme.
Diagnosis: [ICD10] This categorical variable is a patients primary ICD-10 description or also referred to as a patient’s diagnosis. There were 38 categories which were grouped together by the clinical SMEs into seven categories namely:
• [Bipolar] Bipolar disease;
• [GMC] General medical condition;
• [MDD&Anx] Major depressive disorder and anxiety; • [SA] Schizo-affective;
• [S] Schizophrenia;
• [SIPD] Substance induced psychiatric disorder; and
• [Other] which include attention-deficit hyperactivity disorder (ADHD), adjust- ment disorder, dementia, intellectual disability, personality disorders, delu- sional disorder, brief psychotic disorder and intoxication.
Total admissions: This variable is between zero and seven and describes a patient’s total number of admissions during the study period.
Days discharged: [Days] This continuous variable describes the number of days a patient was discharged before the current (re)admissions, if applicable.
Length of stay: [LOS] This continuous variable is the length in days that the patient remained in Stikland Hospital after admission and before being discharged.
Age: The patient’s age on admission stored as a continuous variable.
Substance: C: A binary (0,1) variable indicating whether a patient is reported using (1) or not using (0) cannabis on admission.
Substance: T: A binary (0,1) variable indicating whether a patient is reported using (1) or not using (0) tik on admission.
Substance: A: A binary (0,1) variable indicating whether a patient is reported using (1) or not using (0) alcohol on admission.
Substance: Other: A binary (0,1) variable indicating whether a patient is reported using (1) or not using (0) substances such as mandrax, cocaine, heroin or opioids on admission.
Substance: None: A binary (0,1) variable indicating whether a patient reported not using (1) or using (0) substances on admission.
Another dataset was created that contained grouped information representing ‘lifetime data’. For example, as displayed in Figure 4.3, a patient may have been admitted from Area 2 on his first two admissions, but Area 3 on his third admission, which would then be indicated by a ‘1’ in the column ‘Area: 2’ and ‘Area: 3’. This dataset was more experimental and used for descriptive statistics that may possibly be of interest to the clinical SMEs. Lifetime data for this project refers to the study period (admission(s) between 2012 and 2014).
Figure 4.3: Example of converting the ‘multiple admission entries’ to ‘lifetime data’.
A screenshot taken from Microsoft Excel of the initial cleaned dataset, before the entries were grouped per patient, is displayed in Figure 4.4. The yellow cells indicate that a patient (anonymised folder number) occurs more than once. The substance information in green text indicates that the entry’s substance information is accurate. From this dataset a macro was used to group the entries of recurring patients in one row. The column for each admission was renamed to clearly indicate to which admission the information was applicable to e.g. Age1, Age2, and so forth. This dataset, where a patients occurs only once, was further developed and used in the various analyses.
4.2.
The
Stikland
dataset
81
Figure 4.4: Partial screenshot of the dataset in Microsoft Excel.