How good is New South Wales admitted patient data collection in recording births?
This record linkage study aims to examine the coding concordance of
delivery outcome and discharge status between the New South Wales (NSW)
Midwives Data Collection (MDC) and Admitted Patients Data Collection
(APDC) as well as factors that contribute to hospital births not being
recorded in the APDC. Births recorded in the APDC and MDC datasets for
the calendar year 2005 were used for analysis. Births registered in the
NSW Registry of Births Deaths and Marriages for the same period were
used as validation. Descriptive analysis was used to examine coding
concordance between the APDC and MDC datasets for matched records, and
logistic regression analyses were used to identify factors associated
with hospital births not being included in the APDC. A total of 90,585
unique births were recorded in the MDC for the calendar year 2005. A
total of 79,173 confirmed hospital births were matched to corresponding
records in the APDC; 2,249 (3%) confirmed hospital births were not found
in the APDC. For unmatched records, logistic regression analyses showed
that the level of obstetric hospital in which babies were born was a
significant factor associated with information not being recorded in the
APDC. As compared with local, small isolated, and small metropolitan
hospitals (Levels I to 3 hospitals), larger tertiary hospitals (Levels 4
to 6) and private hospitals had decreased odds of hospital births not
being recorded in the APDC. For matched records, 95% and 99% of records
were found to be coded consistently between the APDC and MDC datasets
for outcome of delivery and discharge status respectively. With a high
level of coding concordance between the APDC and MDC datasets and only a
small percentage of hospital births not being recorded in the APDC, the
obstetrics subset of the APDC dataset was found to be of good quality.
MeSH Keywords: Data Linkage, Data Quality; Birth Records; Medical Records; Clinical Coding; International Classification of Diseases; Hospital Information Systems; Informatics
Supplementary keyword: Health Information Management
Electronic records (Usage)
Information management (Usage)
Medical care (Quality management)
Medical care (Analysis)
|Author:||Lam, Mary K.|
|Publication:||Name: Health Information Management Journal Publisher: Health Information Management Association of Australia Ltd. Audience: Academic Format: Magazine/Journal Subject: Health Copyright: COPYRIGHT 2011 Health Information Management Association of Australia Ltd. ISSN: 1833-3583|
|Issue:||Date: Oct, 2011 Source Volume: 40 Source Issue: 3|
|Topic:||Computer Subject: Information accessibility|
|Geographic:||Geographic Scope: Qatar Geographic Code: 7QATA Qatar|
With the new e-health agenda and the upcoming implementation of the personally-controlled electronic health record (PCEHR), good quality health data and information form an important building block for a good e-health system. For this health information to be useful in the e-health system, data collected at different times in a person's health journey need to be of good quality. There are a number of routine data collections at the population level that may form part of this building block. These include the Admitted Patient Data Collection (APDC) and registries that collect more in-depth data related to specific health conditions that result in hospitalisation. Examples include registries for cancer, birth, perinatal conditions and diabetes. As the use of health information expands, the need to have high quality information becomes more important. It is necessary that the quality of these health datasets are examined and ensured so patients and clinicians can use this information with confidence.
Recent years have seen a number of data quality studies conducted by users of these population health datasets (e.g. Comino et al. 2007; Taylor et al. 2005; McKenzie et al. 2006; Ford et al. 2007; Bell et al. 2008; Hadfield et al. 2008; Roberts, Bell et al. 2008). In terms of the New South Wales (NSW) APDC dataset, with the implementation of data quality check procedures at the Hospital and Health Development level (Australian Bureau of Statistics [ABS] 2008; NSW Department of Health [DoHA] 2009), recent validation studies on obstetric data have shown that data at the unit record level are of good quality (e.g. Lain et al. 2008; Lam et al. 2008; Roberts et al. 2009). However, two remaining data quality issues require further examination: non-inclusion of valid cases, and concordance of coding in health datasets that collect similar or related information. For non-inclusion of valid cases, it is difficult and in most cases unlikely to be identified by applying standard data quality check protocols that examine the data at the unit record level. Non-inclusion could be due to assignment of incorrect International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, Australian Modification (ICD-10-AM) codes, or the incorrect or missed input of information into the hospital information system. Non-inclusion of valid cases may become a potential source of bias for subsequent analysis (Ford, Roberts & Taylor 2006), especially for rare conditions. In the past, the only way to identify the frequency of such occurrence was through medical record audits, a time-consuming and expensive exercise. With advancement in record linkage modelling and algorithms, linkage of a subset of hospital admission data with specific datasets has become an alternative method of data validation (Chen et al. 2010).
In NSW, three data collections are related to births and deliveries: NSW Admitted Patients Data Collection (APDC); NSW Midwives Data Collection (MDC); and birth data from the NSW Registry of Births, Deaths and Marriages (RBDM). The NSW APDC dataset contains data for all inpatient admissions in NSW, which includes demographic and admission related data. For all live births that occur in a hospital an ICD-10-AM code of Z37.0 (single live birth) or Z37.2 (twin, both liveborn) or Z37.3 (twin, one liveborn and one stillborn), or Z37.5 (other multiple, all liveborn) or Z37.6 (other multiple, some liveborn some stillborn) is assigned to the mother's record. For babies delivered at a NSW public hospital, it is DoHA policy that and an ICD-10-AM code in the range Z38.0 to Z38.8 be assigned to the baby's record where the baby is born alive or is born alive and admitted immediately after birth (DoHA 2007). Private hospitals are only required to comply with DoHA client registration policy to the extent required by law (NSW Government 2010); thus, not all babies delivered in private hospitals are assigned a Z38 code.
The NSW MDC records all births in NSW regardless of the baby's place of birth (DoHA 2005:12). Apart from demographic information, this dataset collects all information on maternal health prior to and during pregnancy, the actual pregnancy, labour, delivery, perinatal outcomes, information about previous births, baby's birth characteristics as well as mother and baby discharge status. The birth registration in the NSW RBDM contains information on births registered in NSW. Under the Births, Deaths and Marriages Registration Act 1995 Section 13, a child that is born in the State (or en-route to a port of the State) must be registered within 60 days of birth (NSW Consolidated Acts n.d.). These three datasets can be linked to examine many issues relating to birth and delivery. Figure 1 provides a schematic characterisation of the three datasets and some of the information found in each dataset.
In the hospital admissions data collection, mothers' and babies' records are not linked. Information about the mother's delivery and the baby's birth, as well as other hospital-stay information, are submitted to DoHA as separate records with no information available for the identification of mother and baby dyads within the dataset. This poses difficulties in the examination of coding concordance between mothers' and babies' records when using the APDC datasets alone. The MDC dataset contains information that enables the identification of mother and baby dyads (details about the data linkage process are outlined in the Methods section below). By linking the APDC and MDC records, the mother and baby pairs can be identified, and the coding concordance of mother and baby records can be examined.
By the same token, the APDC, MDC, and RBDM datasets can also be linked to examine the occurrence of non-inclusion of valid cases. The APDC birth and delivery subset contains similar information to the MDC dataset, with the MDC collecting more in-depth delivery and birth information than the APDC for certain data items (DoHA 2005). The MDC collects data on all births, including hospital and non-hospital births. For all births as indicated by the MDC as hospital births, a corresponding record should be found in the APDC dataset. The absence of such records in the APDC dataset can be considered as non-inclusion of valid cases. Therefore, by linking the APDC and MDC datasets one can examine the incidence of non-inclusion, and factors that may contribute to this non-inclusion. For mothers who have delivered a live birth, a birth registration record of the baby should also be found in the RBDM dataset. Thus, the birth records contained in the linked APDC and MDC datasets can be validated by linkage with the RBDM dataset. It is to be noted that cross validation with the ABS 2005 NSW birth counts (ABS 2006) showed that the number of births recorded in RBDM was 10,865 fewer than those recorded in the ABS dataset. The difference found in the two datasets may be due to parents not registering their child within the recommended timeframe. As a result, the RBDM dataset provided a conservative birth validation measure between APDC and MDC.
This study aimed to examine two data quality issues by applying record linkage techniques, linking the birth and delivery subset of the APDC with MDC and birth data from RBDM, focusing specifically on:
1. the percentage of coding concordance between the MDC and APDC for matched records in regards to outcome of delivery and discharge status
2. factors that contribute to hospital birth records not being recorded in the APDC baby dataset (referred to as 'unmatched' records).
Data sources and record selections
Data for babies born between January 1, 2005 and December 31, 2005 were supplied by the data custodians of the APDC, MDC and RBDM datasets. The MDC variable 'baby place of birth' was used to identify all hospital births. The RBDM dataset was used to confirm the birth as recorded in the APDC and MDC datasets.
Record linkage is the process through which records of the same person from different datasets are identified (Centre for Health Record Linkage [CHeReL] n.d.). In NSW, this process is conducted at the CHeReL. The record linkage approach used at CHeReL is similar to 'the best practice protocol' described by Kelman and colleagues (Kelman, Bass & Holman 2002). To ensure patient confidentiality, CHeReL conducts linkage using only identifying data items (i.e. no clinical data). Data custodians provide CHeReL with personal identifiers such as name, address, sex, date of birth, and hospital code together with an encrypted number that refers back to the actual record in the custodian datasets. CHeReL utilises these personal identifiers and applies the probabilistic matching techniques to identify records that are likely to belong to the same person. Once a group of records from different datasets are identified as belonging to the same person, a CHeReL person number is assigned to the group of records for referencing purposes within CHeReL. With the appropriate project ethics approvals, CHeReL then assigns a project person number (PPN) to the group of records that are related to the project. The PPN together with the encrypted number from the data custodians are sent back to the data custodian. Data custodians then extract the approved clinical information from their datasets and forward this information to the researcher together with the corresponding PPNs, which enable researchers to join records for the same person from the different datasets.
Four datasets were supplied by the data custodians (Figure 1). The MDC dataset contained PPNs for both mothers and babies for the required period and the approved data items. The RBDM dataset contained PPNs for the birth. The APDC baby dataset contained PPNs for babies together with approved data items and the APDC mother dataset contained PPNs for mothers and approved information. Using the babies PPNs from the MDC, and the APDC baby datasets, baby records were linked to create two groups of records: hospital births with corresponding APDC records (matched) and hospital births without corresponding APDC records (unmatched). A new indicator variable, 'Match', was created to store this information. Both the matched and unmatched records were linked to the RBDM dataset using the baby PPN to obtain registered birth information. The matched datasets were further linked with the APDC mother dataset to create a large set for the examination of coding concordance between the MDC and APDC datasets. Figure 2 presents a schematic diagram of the linkage process using PPNs supplied by CHeReL and the number of records obtained.
Data management and analysis
SPSS version 17.0 (SPSS Inc. 2007) was used for data management and analysis. Prior to linkage, both MDC and ADPC datasets were checked for duplicates using all variables within the dataset as the criteria. Duplicated records were removed from the dataset and subsequent analysis. PPNs provided by CHeReL were used to link the APDC and MDC datasets. Both matched and unmatched records were cross linked with birth records from RBDM to confirm baby's birth status.
In NSW, apart from the Principal Diagnosis, 54 additional diagnosis fields are also collected. According to the Australian Coding Standards (National Centre for Classification in Health 2004), depending upon the admission condition of the mother and newborn, the delivery related codes (Z37.0 to Z37.9) for the mother and birth related codes (Z38.0 to Z38.8) for the baby can be assigned in any of the 55 diagnosis fields. A data extraction program was written using SPSS to loop through all 55 fields to extract codes in the range of Z37.0 to Z37.9 for mothers and Z38.0 to Z38.8 for babies. The outcome of delivery and birth codes were used for subsequent analyses.
All RBDM confirmed births indicated in the MDC dataset as births occurring in hospital were used in the analysis. For the first aim, matched records were analysed for coding concordance in terms of outcome of delivery and discharge status for both mothers and babies. Two variables were created to examine the concordance of coding in the three datasets. Tables 1 and 2 present the criteria for creating the two concordance variables. A matched record must have the codes and conditions indicated in the tables to be considered as a concordant record. For example, if a mother was assigned a code of Z37.0 (single live birth) in her record, then the corresponding MDC record must either be coded as discharge or transfer or neonatal death (if the baby died during the admission) and have plurality coded as single, and a code of Z38.0 (singleton, born in hospital) must be assigned to the baby record. In the case when a code of Z37.3 (twin, one liveborn, one stillborn) was assigned, then for the live baby, a code indicating discharge or transfer or neonatal death, and a code indicating multiple must be assigned in the MDC, and a code of Z38.3 (twin, born in hospital) must be assigned to the baby record. For the stillborn twin, codes indicating stillborn and multiple births must be coded in the MDC, and there must be no record in the APDC baby dataset for the stillbirth. For concordance coding of discharge status across the three datasets (Table 2), if a mother was assigned a discharged code in the MDC dataset, then she must also be assigned one of the discharge codes (discharged on leave, discharged by hospital, discharged at own risk) in the APDC mother datasets for the records to be considered in accord. The same principle applied to the other discharge categories for both mothers and babies. Descriptive analyses were conducted on the two concordance variables.
[FIGURE 2 OMITTED]
For the second aim of the project (factors associated with hospital births not recorded in APDC) the indicator variable, 'Match', was used as the dependent variable for the analysis. Independent factors under examination were level of obstetric hospital; mother and baby characteristics; labour, presentation, delivery characteristics; mother and baby separation status; and baby conditions. Level of obstetric hospital was grouped into the categories: Levels 1 to 3 (local, small isolated, small metropolitan hospitals), Levels 4 to 6 (large metropolitan, tertiary hospitals) and private hospitals. According to definition provided in the NSW Health Mothers and Babies report:
* Level 1 hospitals are local hospitals (no birth), with postnatal only facilities.
* Level 2 hospitals are small isolated hospitals, responsible for low-risk births only and staffed by general practitioners and midwives.
* Level 3 hospitals are country district and smaller metropolitan hospitals, care for mothers and infants at low-moderate risk, with full resuscitation and theatre facilities, and staffed by rostered obstetricians, resident medial staff, midwives, accredited general practitioners and on call specialist anaesthetist, has level 2b neonatal care (a unit that can give low-level oxygen and has a paediatrician on call).
* Level 4 hospitals are country base-metropolitan district hospitals, delivery and care for mothers and/or babies with moderate risk, staff by obstetricians, paediatrician and rostered resident medical staff and on call specialist anaesthetist, has level 2b neonatal care.
* Level 5 hospitals are country base-metropolitan district hospitals, with facilities to care for high risk mothers and infants, has level 2a neonatal care (a unit that can give high-level oxygen, can start mechanical ventilation if necessary and has paediatric house staff).
* Level 6 hospitals are tertiary specialist obstetrics hospitals able to care for low to high risk births, has level 3 neonatal intensive care. (DoHA 2005:16)
Mothers' characteristics included the mother's age ([greater than or equal to] 40 or <40); whether she had maternal diabetes, maternal hypertension, gestational diabetes or pre-eclampsia. The cut-off for mothers' age was chosen because research has shown that there is a relationship between maternal age and foetal loss with mothers aged 40 or above having a higher chance of foetal loss (Nybo Andersen et al. 2000). Babies' characteristics included sex of baby; plurality (singleton or multiple); order of birth (first or others). Labouring was categorised as spontaneous, induced, no labour; presentation was categorised into vertex, breech and other; delivery was categorised into normal vaginal, vacuum extraction, vaginal breech, forceps and caesarean section. Mothers' separation status was categorised as discharged and transferred/others; babies' separation status was categorised as discharged, stillborn/neonatal death, transferred. Babies' conditions included gestational age (<37 weeks, 37-41 weeks and >41 weeks) and birth weight (above 1.5 kg and below 1.5 kg).
Logistic regression was used to determine factors associated with hospital births, as indicated in the MDC, not being recorded in the APDC. Unadjusted logistic regressions were conducted to examine bivariate associations between all independent factors and the dependent variable. Data were then subjected to multivariate analysis to determine the adjusted association between independent factors and whether hospital births were recorded in the APDC. For the inclusion of any variable in the multivariate analysis, the criterion of a bivariate association with p<.05 was used. A backward model selection approach was used. Non-significant independent variables were progressively dropped from the model, with variables with the highest p-value eliminated first. The elimination process continued until all remaining variables were significant at .001 level (two-sided). Due to the large sample size, any slight variation is likely to become significant. In order to avoid an inflated Type I error rate, a significance level of .001 was chosen for this analysis.
Figure 2 presents the number of records found in the APDC and MDC datasets and the selection procedure for the final analysis datasets. A total of 90,610 birth records for the calendar year 2005 were recorded in MDC. A slightly smaller number of records were recorded in APDC (89,213) for the same period. After the removal of duplicate records in each dataset, 90,583 and 88,624 unique records were found in the MDC and APDC datasets respectively. Linking the two datasets using PPNs provided by CHeReL, yielded a total of 87,408 linked records. There were 3,177 (3.5%) MDC records with no corresponding APDC match. Of these unmatched records, 2,428 (76.4%) records were recorded in the RBDM birth dataset as confirmed births. Of the confirmed birth records that were not included in the APDC dataset, 2,249 (92.6%) were births that occurred in hospitals according to the MDC place of birth classification. Of matched records, 79,173 (90.6%) records were confirmed births that occurred in hospitals according to the MDC. All matched records were found to have a corresponding mother record in the APDC mother dataset.
Table 3 shows the percentage concordance in coding for delivery outcome and discharge status obtained from the matched data among the APDC mother, MDC and APDC baby datasets. A high level of concordance was found with the matched set with a 95.6% (N=75,692) concordance for the coding of outcome of birth across the three datasets, and a 99.1% (N=78,425) concordance for the coding of discharge status.
Table 4 provides bivariate associations between the dependent variable and independent factors: level of obstetric hospital; mother and baby characteristics; labour, presentation, delivery characteristics; mother and baby separation status; and baby conditions and the unmatched records. Significant (p<.05) unadjusted associations were found between the dependent variable and level of obstetric hospital; maternal age, maternal diabetes and gestational diabetes; labour, presentation and delivery characteristics; mother and baby separation status; plurality and birth order; as well as gestation age and birth weight of baby. Hence, these variables were included in further regression analyses.
Table 5 gives results (odds ratios and confidence intervals [CI]) obtained from the multivariate logistic regression analysis. Results indicated that some birth variables for the baby and one of the hospital variables were significantly associated with hospital birth records not being recorded in the APDC (unmatched). These included: type of hospital in which the baby was born, baby discharge status and gestational age. The odds of unmatched records was also increased for premature birth (<37 weeks) with unmatched records being more than two times as likely to be premature babies in comparison to full term babies (OR=2.25, 95%CI=1.942.61). The most significant variable associated with unmatched records was the discharge status of the baby. The odds of records being unmatched were more than 100 for babies classified as stillborn or neonatal death when compared to the normal discharge (OR=106.92, 95%CI = 80.15-142.63). Conversely, likelihood of unmatched records was reduced for types of hospital. There was a reduction of odds of unmatched records of nearly 70% and 30% for Levels 4-6 and private hospitals respectively (Levels 4-6 hospital: OR=0.32, 95%CI=0.28-0.36; private: OR=0.74, 95%CI=0.66 0.83).
Using linkage of related public health datasets, this study examined two data quality issues relating to hospital birth records: concordance of codes assigned across the APDC mothers, APDC babies and MDC datasets for birth outcome and discharge status; and factors associated with NSW hospital birth records not being recorded in the NSW APDC. Using the NSW MDC categorisation of baby's place of birth and the birth registration in the NSW RBDM as the standard in identifying registered hospital births, it was found that only 3% of registered hospital births did not have a corresponding record in the APDC baby dataset. For the matched dataset, the coding concordance for outcome and discharge status was very high across the three birth related datasets, with above 95% concordance for both. In terms of factors associated with hospital birth not being recorded in the APDC, it was found that the type of hospitals in which the baby was born, the baby's separation status and gestational age were significantly associated with hospital birth not being recorded in APDC.
This study's finding of a high level of coding concordance with respect to birth outcome and discharge status in the matched data is consistent with past findings. Past research examining different aspects of obstetric coding in the APDC dataset has found that overall data quality is good. For example, researchers have found a high level of reliability of recording caesarean birth (Chen et al. 2010); obstetric haemorrhage (Lain et al. 2008); general anaesthesia used for childbirth (Roberts, Ford et al. 2008). Future analysis could further investigate the characteristics of the 4.4% records that were found to have non-concordant birth outcome coding between the
APDC and MDC datasets.
Using logistic regression to examine the determinants of data quality (e.g. factors associated with errors of omission such as hospital births not being recorded in the APDC baby dataset) is a relatively new analytical method. A recent study conducted by Chen and colleagues (2010) is the only other study that utilised this analysis methodology in the examination of factor association with the misclassification of caesarean section at last birth in birth records. Analysis carried out in this study found that the factors that were significantly associated with hospital births not being recorded in the APDC baby datasets were babies that were stillborn or born at a low gestational age, and those born in local, small isolated, and small metropolitan hospitals (Level 1 to 3 hospitals). Babies born with a low gestational age were less likely to have survived. According to the NSW Client Registration Guideline (DoHA 2007), babies born alive in public hospitals and some private hospitals (according to the admission requirements of the private hospital) will be assigned a new medical record and outcome code in the range of Z38.0 to Z38.8. For babies that did not survive, only a code Z37.1 (Single stillborn); or Z37.3 (Twin, one liveborn and one stillborn); or Z37.4 (Twin, both stillborn); or Z37.6 (Other multiple, some liveborn) or Z37.7 (Other multiple, all stillborn) was recorded in the mother's record. Thus it was indeed correct that there should be no APDC baby records for babies who were stillborn.
One interesting finding from this study is that, when compared to babies who were born in Levels 1 to 3 hospitals, babies born in Levels 4 to 6 hospitals and private hospitals had decreased odds of hospital births not being included in the APDC. This finding requires further investigation. Future research should examine factors that may contribute to the smaller Levels 1 to 3 hospitals being more likely than larger public and private hospitals to have hospital births not recorded in the APDC baby dataset.
While this study has linked the matched set with the APDC mother dataset to cross validate the three birth-related datasets, it has not linked the unmatched set with the APDC mother dataset to further validate what the mother's APDC record indicated as the outcome of delivery. If the mother record indicated an ICD-10-AM code of Z37.0 (Single live birth), Z37.2 (Twin, both liveborn), Z37.3 (Twin, one liveborn and one stillborn), Z37.5 (Other multiple, all liveborn), or Z37.6 (Other multiple, some liveborn) then a corresponding record should be found in the APDC baby dataset. A missing APDC baby record from the dataset is indeed an error. Further analysis should be conducted to examine the possible reasons for these records not being recorded.
The hospital obstetric dataset is of good quality with only a small percentage of hospital births not being recorded in the APDC baby dataset and the matched datasets showing a high level of concordance in the coding of delivery outcome and discharge status. With the new e-health agenda and the upcoming implementation of PCEHR, the NSW APDC obstetric subset provides reliable information for this important building block for a good e-health system.
This project has been approved by the University of Sydney Human Research Ethics and NSW Population and Health Service Research Ethics Committees.
Australian Bureau of Statistics (2008). NSW Health Department Admitted Patient Data Collection, February 2008 .Available at: http://www.abs.gov.au/AUSSTATS/abs@.nsf/Lookup/Explanat ory%20Notes1452007?opendocument&tabname=Notes&prod no=1368.1&issue=2007&num=&view (accessed 8 December 2010).
Australian Bureau of Statistics (2006). Birth Australia 2005, October 2006. Available at: http://www.abs.gov.au/AUSSTATS/abs@. nsf/DetailsPage/3301.02005?OpenDocument (accessed 5 January 2011).
Bell, J. C., Ford, J. B., Cameron, C. A. and Roberts, C. L. (2008). The accuracy of population health data for monitoring trends and outcomes among women with diabetes in pregnancy. Diabetes Research and Clinical Practices 81(1): 105-109.
Centre for Health Record Linage (n.d.). Guide to health record linkage services, version 1.3. Available at: http://www.cherel.org. au/GuidetoCHeReLservices.pdf (Accessed 1 November 2010).
Chen, J. S., Roberts, C. L., Ford, J. B., Taylor, L. K. and Simpson, J. M. (2010). Cross-sectional reporting of previous Cesarean birth was validated using longitudinal linked data. Journal of Clinical Epidemiology 63(6): 672-678.
Comino, E. J., Titmuss, A., Harris, E., Craig, P. and Pulver, L. J. (2007). Identification of Aboriginal infants at an urban hospital. Journal of Paediatrics and Child Health 43(9): 623-626.
Ford, J. B., Roberts, C., Taylor, L. K. (2006). Characteristics of unmatched maternal and baby records in linked birth records and hospital discharge data. Paediatric and Perinatal Epidemiology 20(4): 329-337.
Ford, J. B., Roberts, C. L., Algert, C., Bowen, J. R., Bajuk, B. and Henderson-Smart, D. R. (2007). The validity of hospital discharge data for determining neonatal morbidity and mortality. BMC Health Services Research 7: 188.
Hadfield, R. M., Lain. S, J., Cameron, C. A., Bell, J. C., Morris, J. M. and Roberts, C. L. (2008). The prevalence of maternal medical conditions during pregnancy and a validation of their reporting in the hospital discharge data. Australian and New Zealand Obstetrics and Gynaecology 48(1): 79-82.
Kelman, C. W., Bass, A. J. and Holman, C. D. J. (2002). Research use of linked health data--a best practice protocol. Australian and New Zealand Journal of Public Health 26(3): 251-255.
Lain, S. J., Roberts, C. L., Hadfield, R. M., Bell, J. C. and Morris, J. M. (2008). How accurate is the reporting of obstetric haemorrhage in hospital discharge data? A validation study. Australian and New Zealand Journal of Obstetrics and Gynaecology 48(5): 481-484.
Lam, M., K., Innes, K., Saad, P, Rust , J., Dimitropoulos, V and Cumerlato, M. (2008). An evaluation of the quality of obstetric morbidity coding using an objective assessment tool, the Performance Indicators For Coding Quality (PICQ). The Health Information Management Journal 37(2): 19-29.
McKenzie, K., Harding, L. F., Walker, S. M., Harrison, J. E., Enraght-Moony, E. L. and Waller, G. S. (2006). The quality of national data on injuries requiring hospitalisation. Medical journal of Australia 185(9): 185.
National Centre for Classification in Health (2004). International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, Australian Modification, 4th edition. Sydney, National Centre for Classification in Health, University of Sydney.
NSW Department of Health [DoHA] (2005). NSW Public Health Bulletin Supplement: NSW Mothers and Babies 2005, 18(S1). Sydney, NSW Department of Health.
NSW Department of Health [DoHA] (2007). Client Registration Guideline. Available at: http://www.health.nsw.gov.au/policies/ gl/2007/pdf/GL2007_024.pdf (accessed 10 December 2010).
NSW Department of Health [DoHA] (2009). NSW Health Services Comparison Data Book 2007/2008 Volume 1, Sydney, NSW Department of Health.
NSW Government (2010). NSW Government Legislation: Private Health Facilities Regulation 2010. Available at: http://www.legislation.nsw.gov.au/maintop/view/inforce/ subordleg+64+2010+cd+0+N (accessed 6 December 2010).
NSW Consolidated Acts (n.d.). Births, Deaths and Marriages Acts 1995--Births Registration. Available at: http://www.austlii. edu.au/au/legis/nsw/consol_act/bdamra1995383/s13.html (accessed 1 December 2010).
Nybo Andersen, A. M., Wohlfahrt, J., Christens, P, Olsen, J. and Melbye, M. (2000). Maternal age and fetal loss: population based register linkage study. BMJ 320(7251): 1708-1712.
Roberts, C. L., Bell, J. C., Ford, J. B., Hadfield, R. M., Algert, C. S. and Morris, J. (2008). The accuracy of reporting of the hypertensive disorders of pregnancy in population health data. Hypertension in Pregnancy 27(3): 285-297.
Roberts, C. L., Bell, J. C., Ford, J. B. and Morris, J. (2009). Monitoring the quality of maternity care: how well are labour and delivery events reported in population health data? Paediatric and Perinatal Epidemiology 23(2): 144-152.
Roberts, C. L., Ford, J. B., Lain, S., Algert, C. S. and Sparks, D. J. (2008). The accuracy of reporting of general anaesthesia for childbirth: a validation study. Anaesthesia and Intensive Care 36(3): 418-424.
SPSS Inc (2008). SPSS Base V.17.0 for Windows. Chicago, IL, SPSS Inc.
Taylor, L. K., Travis, S., Pym, M., Olive, E. and Henderson-Smart, D. J. (2005). How useful are hospital morbidity data for monitoring conditions occurring in the perinatal period. Australian and New Zealand Journal of Obstetrics and Gynaecology 45(1): 36-41.
Mary K Lam PhD
Faculty of Health Science
The University of Sydney
PO Box 170
Lidcombe NSW 1825
Tel: +61 2 9351 9570
Table 1: Criteria for concordance of birth outcome across the APDC mother, MDC and APDC baby datasets APDC MOTHERS MDC ICD-I0-AM CONDITION BABY DISCHARGE STATUS PLURALITY CODE * Z37.0 AND Discharge or Transfer or died 1 Z37.1 AND Stillborn 1 Z37.2 AND Discharge or Transfer or died 2 Z37.3 AND Discharge or Transfer or died 2 Stillborn 2 Z37.4 AND Stillborn 2 Z37.5 AND Discharge or Transfer or died >2 Z37.6 AND Discharge or Transfer or died >2 Stillborn >2 Z37.7 AND Stillborn >2 APDC MOTHERS APDC BABIES ICD-I0-AM CONDITION ICD-I0-AM CODE * CODE ** Z37.0 AND Z38.0 Z37.1 AND no APDC baby record Z37.2 AND Z38.3 Z37.3 AND Z38.3 no APDC baby record Z37.4 AND no APDC baby record Z37.5 AND Z38.6 Z37.6 AND Z38.6 no APDC baby record Z37.7 AND no APDC baby record * Z37.0--Single live birth; Z37.I--Single stillbirth; Z37.2--Twin, both liveborn; Z37.3--Twin, one liveborn one stillborn ; Z37.4--Twin both stillborn; Z37.5--Other multiple all liveborn; Z37.6--Other multiple, some liveborn some stillborn; Z37.7--Other multiple, all stillborn ** Z38.0--Singleton, born in hospital; Z38.3--Twin, born in hospital; Z38.6--Other multiple, born in hospital Table 2: Criteria for concordance of discharge status MDC APDC SEPARATION MODES CONDITION SEPARATION MODES Mothers Discharged AND Discharge on leave OR Discharge by hospital OR Discharge at own risk Died AND Died (autopsy) OR Died (no autopsy) Transferred AND Transfer to nursing home OR Transfer to psych hospital OR Transfer to other hospital OR Transfer to other accommodation OR Transfer to palliative care unit/hospice Babies Discharged AND Discharge on leave OR Discharge by hospital OR Discharge at own risk Stillborn or AND Died (autopsy) OR Died (no Neonatal death autopsy) Transferred AND Transfer to nursing home OR Transfer to psych hospital OR Transfer to other hospital OR Transfer to other accommodation OR Transfer to palliative care unit/hospice Table 3: Coding concordance between MDC, APDC mother and APDC baby datasets for matched records in regards to delivery outcome and discharge status CONCORDANCE FREQUENCY PERCENTAGE Delivery Outcome 75692 95.6 Discharge Status 78425 99.1 Table 4: Frequencies, percentages, and unadjusted odds ratios, 95% confidence interval (CI) for the characteristics of independent factors by matched or unmatched MDC records CHARACTERISTICS FREQUENCY UNADJUSTED % RESULTS Unmatched Matched (n-2249) (n=79I73) Level of obstetric hospital * 4/6 1022 (45.4) 48793 (61.6) Private 739 (32.9) 20438 (25.8) 1/3 ** 488 (21.7) 9942 (12.6) Mother's characteristics Age * 240 105 (4.7) 2906 (3.7) <40 ** 2144 (95.3) 79262 (96.3) Maternal hypertension Yes 24 (1.1) 781 (1.0) No ** 2225 (95.9) 78392 (99.0) Gestational diabetes * Yes 2167 (96.4) 75267 (95.1) No ** 82 (3.6) 3906 (4.9) Pre-eclampsia Yes 2138 (95.1) 74913 (94.6) No ** 111 (4.9) 4260 (5.4) Labour, presentation and delivery Labour * Spontaneous ** 1116 (49.6) 44829 (56.6) Induced--all 696 (30.9) 20559 (26.0) No labour 437 (19.4) 13772 (17.4) Presentation * Vertex ** 1996 (89.0) 74857 (94.7) Breech 214 (9.5) 3632 (4.6) Other 32 (1.4) 561 (0.7) Delivery Normal vaginal ** 1204 (53.5) 46763 (59.1) Vacuum extraction 169 (7.5) 5869 (7.4) Vaginal breech 107 (4.8) 282 (0.4) Forceps 70 (3.1) 2563 (3.2) Ceasarean section 699 (31.1) 23696 (29.9) Separation status Mother * Discharged ** 2135 (95.1) 76445 (96.6) Transferred/others 111 (4.9) 2708 (3.4) Baby * Discharged ** 1669 (74.2) 74908 (94.6) Stillborn/ neonatal death 419 (18.6) 163 (0.2) Transferred 160 (7.1) 4075 (5.1) Baby's characteristics Sex Male ** 1116 (49.8) 40755 (51.5) Female 1124 (50.2) 38334 (48.5) Plural Singleton ** 2111 (93.9) 76610 (96.8) Twin/triplets 138 (6.1) 252563 (3.2) Parity * 1st ** 2179 (96.9) 77878 (98.4) Others 70 (3.1) 1295 (1.6) Baby's conditions Gestation age * Full term (37-41 wks) ** 1642 (73.0) 72546 (91.6) Preterm birth (<37 wks) 578 (25.7) 5178 (6.5) Post term (>41 wks) 28 (1.2) 1443 (1.8) Birth weight * Above 1.5 kg ** 1948 (87.9) 78488 (99.2) Below 1.5 kg 268 (12.1) 670 (0.8) CHARACTERISTICS Odds 95% CI ratio Level of obstetric hospital * 4/6 .43 .38-.48 Private .74 .66-.83 1/3 ** 1.00 - Mother's characteristics Age * 240 1.29 1.05-1.57 <40 ** 1.00 - Maternal hypertension Yes 1.08 .72-1.63 No ** 1.00 - Gestational diabetes * Yes .73 .58-.91 No ** 1.00 - Pre-eclampsia Yes .91 .75-1.11 No ** 1.00 - Labour, presentation and delivery Labour * Spontaneous ** 1.00 -Induced--all 1.36 1.24-1.50 No labour 1.28 1.14-1.43 Presentation * Vertex ** 1.00 -Breech 2.21 1.91-2.55 Other 2.14 1.49-3.06 Delivery Normal vaginal ** 1.00 -Vacuum extraction 1.12 .95-1.32 Vaginal breech 14.74 11.71-18.54 Forceps 1.06 .83-1.35 Ceasarean section 1.15 1.04-1.26 Separation status Mother * Discharged ** 1.00 -Transferred/others 1.47 1.21-1.78 Baby * Discharged ** 1.00 -Stillborn/ neonatal death 115.37 95.66-139.14 Transferred 1.76 1.49-2.08 Baby's characteristics Sex Male ** 1.00 -Female 1.07 .99-1.17 Plural Singleton ** 1.00 -Twin/triplets 1.95 1.64-2.33 Parity * 1st ** 1.00 -Others 1.93 1.51-2.47 Baby's conditions Gestation age * Full term (37-41 wks) ** 1.00 -Preterm birth (<37 wks) 4.93 4.47-5.44 Post term (>41 wks) .86 .59-1.25 Birth weight * Above 1.5 kg ** 1.00 -Below 1.5 kg 16.12 13.89-18.70 * Signficant factors (p<.05) ** Reference group Table 5: Unadjusted and adjusted odds ratios and 95% confidence intervals (CI) of significant independent factors in the final logistic regression model UNADJUSTED Significant factors Odds ratio 95% CI Level of obstetric hospital 4/6 .43 .39-.48 Private .74 .66-.83 1/3 ** 1.00 - Baby separation status Stillborn/neonatal death 115.38 95.66-139.15 Transferred 1.76 1.50-2.08 Discharged ** 1.00 - Gestation age Preterm (<37 wks) 4.93 4.47-5.44 Post term (>41 wks) .86 .59-1.25 Full term (37-41 wks) ** 1.00 - ADJUSTED Significant factors Odds ratio 95% CI Level of obstetric hospital 4/6 .32 .28-.36 Private .74 .66-.83 1/3 ** 1.00 - Baby separation status Stillborn/neonatal death 106.92 80.15-142.63 Transferred 1.74 1.46-2.07 Discharged ** 1.00 - Gestation age Preterm (<37 wks) 2.25 1.94-2.61 Post term (>41 wks) .98 .66-1.44 Full term (37-41 wks) ** 1.00 - ** Reference group Figure 1: Schematic characterisation of the NSW APDC mother, APDC baby, MDC and RBDM datasets, and examples of some of the variable/information available in each of the datasets. PPNs are generated by the CHeReL and are used by the researcher to link the different datasets. RBDM--BIRTH * PPN-BABY MDC--MOTHERS & APDC MOTHERS BABIES APDC BABIES * PPN-mother * PPN-mother * PPN-baby * PPN-baby * Mother * Mother pregnancy * Baby characteristics characteristics characteristics * Separation status * Baby * Separation status * ICD-I0-AM characteristics * ICD-I0-AM diagnoses * Birth and delivery diagnoses codes (provides characteristics codes (provides information about * Mother and baby information about outcomes of separation status the birth) delivery) * Procedures codes * Procedures codes
|Gale Copyright:||Copyright 2011 Gale, Cengage Learning. All rights reserved.|