Better health and ageing for all Australians

Evaluation of the mental health nurse incentive program

Appendix E: MHNIP cost analysis: description and findings

Up to Publications

prev page| TOC |next page

Introduction
Method
Results
Discussion

Introduction

The statement of requirement for the RFQ called for an analysis of the cost benefits. Specifically;
  • overall cost benefit of the MHNIP on the health system and drivers of this benefit; and
  • overall cost and benefits on the health system of extending the program to the private hospital setting.
The term 'cost-benefit' is used in this report in the generic or commonly used interpretation as a comparative study of the benefits and costs using a combination of qualitative and quantitative measures.i HMA considered the appropriateness of conducting a cost-effectiveness analysis (CEA) of MHNIP. However, the incomplete resource data available and great uncertainties in capacity to estimate outcomes in economically relevant units were such that CEA was deemed inappropriate. It was on this basis that a cost analysis was conducted.

This appendix presents the results of the cost-analysis. The cost analysis focusses on the level of resource use in treating patients with a severe mental illness under the MHNIP service delivery model and in the absence of MHNIP services. Therefore, the study assesses the change in key-resource use of patients with a severe mental illness receiving services under MHNIP and those patients with a severe mental illness that do not receive services under MHNIP. The study employs a retrospective longitudinal study design, where patients involved in the study are their own comparator.
Top of page

Method

As part of the Evaluation of the Mental Health Nurse Incentive Program, 18 case studies were conducted across Australia. These case studies provided a wealth of both qualitative and quantitative information. In addition to consulting with a number of people at each of these sites, HMA also sought de-identified information on up to 50 consumers of MHNIP services at each case study organisation. For simplicity, organisation were requested to select their last 50 MHNIP consumers. Information was high-level in nature and included information on:
  • age;
  • sex;
  • entry date (to MHNIP);
  • exit date (from MHNIP, if relevant);
  • HoNOS scores; and
  • hospitalisations – 12 months prior to joining MHNIP and 12 months after entry into MHNIP.
Where possible, the site was asked to record the principle diagnosis (or reason for admission), admission/separation dates, along with length of stay (LOS) of the reported hospitalisations. The Diagnostic Related Group (DRG) for the admission was also requested, but was not expected to be known by the eligible organisation.

Two case study sites were not required to produce de-identified patient information. One service had previously provided services under MHNIP but had since ceased. The other organisation had only begun providing MHNIP services a short time prior to the case study visit. HMA received completed templates from 15 of the possible 16 case study sites. This represented de-identified patient information on 464 consumers of MHNIP services. Caution should be used when generalising the results of this cost analysis to the overall MHNIP as this sample size is not representative. A much larger sample size should be used to inform a more comprehensive economic analysis.

A pre-determined inclusion criterion was applied to each of the patients and hospitalisations. Records from two organisations were wholly excluded from the analysis on the basis that information on hospitalisations related to the patient's whole of life, rather than only the 12 months prior to entry into MHNIP (112 patients). These hospitalisations were unable to be classified into the correct 12-month period, as no dates were given. Patients were also excluded from analysis where the patient had entered MHNIP less than a year ago (after 1 September 2011) and had not yet exited the program (84 patients). Those patients who entered MHNIP after 1 September 2011, but had since exited the program were included in the analysis (31 patients). These patients were included in the analysis based on the premise that exiting the MHNIP signalled that they were not at risk of hospitalisation. One patient was excluded on the basis that the patient did not have an entry or exit date for the MHNIP.

A total of 267 patients included in our analysis recorded 34 hospitalisations in the 12 months prior to entering MHNIP, and 30 hospitalisations in the 12 months after entering the MHNIP. More complete information was available for hospitalisations that occurred in the 12 months after entering MHNIP, with a 90.0% completion rate for LOS (n=25). Length of stay was complete for 73.5% (n=25) of the hospitalisations that occurred in the 12 months prior to entry to MHNIP.

An expected length of stay was assigned to each of the hospitalisations missing LOS based on the average length of stay for patients with that primary mental health diagnosis in that 12-month period.

Primary and secondary mental health diagnoses, which were represented by open text fields in the data collection template, were coded into the major mental health diagnoses as presented in Tolkien IIii. A number of mental health diagnoses presented did not fall neatly into this structure such as adjustment disorder, personality disorder (unspecified), organic personality disorder, schizoaffective disorder and postnatal depression. These primary mental health diagnoses were not coded to the Tolkien II structure and were left as their own distinct categories. Three sub-categories of anxiety disorders (panic/agoraphobia, social phobia and generalised anxiety disorder) were combined into one category (Anxiety) due to the low specificity and completeness of the raw data.

Information on the pattern of MBS Item claims for MHNIP patients in the 12 months before and after entering MHNIP was unavailable. The total number of claims for Medicare Item Numbers 2710 and 2712 by MHNIP patients (regardless of when they entered MHNIP) could not be accessed for this study (this was beyond the scope of the evaluation ethics approval).
Top of page

Results

Analysis of the de-identified patient information was supplemented by the qualitative information collected as part of the case study and survey processes. Many of the identified potential differences in costs were unable to be quantitatively measured and are discussed below.

MHNIP sessional payments

The total number of MHNIP sessions dedicated to a patient in the 12 months following entry to MHNIP was not quantitatively measured and would be difficult to quantitatively measure on a retrospective basis. Feedback from eligible organisations participating as a case study site indicated that MBS Claim Forms did not reliably measure the number of patients seen within a session. Several mental health nurses interviewed during the case studies advised they claimed a maximum of two patients per session, regardless of how many patients they actually supported during a session. The reason given was that any additional information above two patients was not of relevance to the organisation; funding was not affected and it reduced the administrative and data input requirements by the mental health nurse. This was also an observation reported by the ACMHN.

Many of the case study sites indicated that the average time spent face-to-face with patients during a session was approximately one hour. The frequency of contact with the mental health nurse varied greatly for those patients spoken to as part of the case study visits (up to five consumers per case study site). Patients were asked how often they currently see the mental health nurse. While this varied from twice a week, to once every 6 months, many indicated that at first they saw the mental health nurse weekly (and in some cases more often), but had moved to less frequent appointments as their condition improved.

From the case studies it was determined that a common frequency of contact by patients with their mental health was approximately one hour every week for the first six months following entry into MHNIP and this moved to fortnightly appointments thereafter. A contact profile of this frequency consumes 39 hours of the mental health nurse's time in the first 12 months after entry to MHNIP. This implies that a total of 11.1 sessions were dedicated, on average, to each consumer in this period. Using these estimates, the cost of providing MHNIP services ranged from $2,674 for consumers attending metropolitan practices, to $3,343 for those located in non-metropolitan areas in the 12 months following entry to MHNIP.

More detailed information on the level of service provision to patients in the 12 months following entry to MHNIP will need to be collected prospectively to place any certainty around these estimates.

MBS items claimed

While detailed data on MBS items claimed on behalf of MHNIP patients was not available, anecdotal feedback provided during case study visits suggested that medical practitioners had shorter consultations with patients since joining MHNIP. Other medical practitioners indicated that patients receiving services under MHNIP were less likely to have unscheduled visits. For scheduled appointments, many medical practitioners said the actual duration of consultations was more closely aligned to the scheduled appointment duration.

Attributing any changing patterns in MBS Items claimed on behalf of MHNIP patients to MHNIP was difficult given that patients see their GPs for other medical conditions unrelated to their mental health. The over-all effect on MBS Items claimed in the two periods is ambiguous and should be further explored in future analysis when data is available.

Pharmaceuticals

Detailed information on the use of pharmaceuticals by patients receiving services under the MHNIP was not available. Feedback provided during the case study visits indicated that the activities of the mental health nurse improved compliance and contributed significantly to the management and monitoring of medication for patients. However, measuring changes in pharmaceutical spending (through the PBS) for MHNIP patients may not be appropriate, as there will always be a cohort of consumers that require medication as part of the management of their condition. Furthermore, increased pharmaceutical use may also be clinically appropriate. A relevant outcome relating to pharmaceuticals under MHNIP is increased compliance and better management of medication, rather than a reduction in the use of pharmaceuticals. Improved compliance and better management of medications will result in better patient outcomes, and perhaps a reduction in the aggregate HoNOS score.
Top of page

Hospitalisations avoided

A summary of hospitalisations by primary health diagnosis is provided in Table 8.5. In the 12 months prior to entry into MHNIP patients had an average hospital length of stay of 4.74 days (95% CI 2.18 – 7.30). The average length of stay was reduced to 1.99 days (95% CI 0.74 – 3.25) for this same group of patients in the 12 months after entering MHNIP. This implies an average reduction in hospital length of stay of 2.75 days per patient (95% CI -5.58 – 0.078). A paired t-test was used to confirm that the reduction in average length of stay per patient was statistically significant at the 0.10 level (p = 0.058).

Those with a primary mental health diagnosis of schizophrenia reported the greatest reduction in average length of stay (mean -20.42 days; 95% CI -37.74 – -3.10). While only 6 of the 36 patients with schizophrenia in the sample were hospitalised for a mental health related condition in the 12 months prior to entering MHNIP (each of these cited schizophrenia as the primary reason for being admitted), the time that they spent in hospital was considerably longer (mean = 126 days; median 137 days; 95% CI 78.36 – 173.64 days) than for patients with other primary mental health diagnoses.

Patients with other primary mental health diagnoses, such as anxiety, depression and personality disorder (unspecified), also reported a reduction in the average length of hospital stay per patient. However, these results were not statistically significant (p > 0.10). It is likely that the small sample size of each of these sub-groups contributed to this variability.

Additional sub-group analysis was conducted on the basis of age and gender, but no statistically significant differences between sub-groups were present. Given the small number of patients within most of these sub-groups, this finding was not surprising.

An average per diem cost for each of the relevant DRGs were retrieved from Round 14 (2009-2010) of the National Hospital Cost Data Collection (NHDC) Cost Weights for AR-DRG Version 6.0x (Public Hospitals)iii. Each hospitalisation was assigned to a DRG based on the reported principle diagnosis of that hospitalisation. For those hospitalisations that did not have a principle diagnosis (n=11), a DRG was assigned based on the patient's primary mental health diagnosis. On average, the expected cost of hospitalisations fell from $4,418 pre MHNIP intervention (95% CI $1,449 – $7,387) to $1,998 post MHNIP intervention (95% CI $746 – $3,250). A paired t-test indicated that the reduction in the expected cost of hospitalisation was again statistically significant at the 0.10 level (p=0.066).

A uniform per diem cost was also derived from the NHCDC Cost Weights by weighting the average per diem cost of a range of DRGs related to mental healthiv by the total length of stay for that DRG. This resulted in an average per diem cost of $960 per patient. The hypothesised savings per patient from this analysis closely approximated the results under the scenario reported above where hospitalisations were mapped to a DRG. The expected cost of hospitalisation per patient prior to joining MHNIP was $4,551 (95% CI $2,093 – $7,011). The expected cost of hospitalisation in the 12 months following entry to MHNIP was $1,912 (95% CI $706 – $3,120) per patient, a reduction of $2,639 (95% CI -$5,353 – $75). A paired t-test confirmed that this reduction was also statistically significant at the 0.10 level (p=0.058).

Caution should be taken when interpreting these savings, given the large confidence intervals. The variability of these results is the product of a small sample size, and the low rates of hospitalisations for these patients.

Presentations to emergency departments

Discussions at case study sites indicated that MHNIP patients presented less frequently to hospital emergency departments than they did prior to receiving services under the program. The exact number of attendances to emergency departments for each consumer was unknown, and unlikely to be reliably recorded on a retrospective basis. Further investigation into ED attendances for this patient cohort could be considered.

Patient outcomes

requirement of MHNIP is that a HoNOS measure should be completed every 90 days for patients receiving services under MHNIP. HoNOS scores were not recorded uniformly within our patient sample, with many patients having 'missing' HoNOS scores at different points since entering MHNIP. HoNOS scores were not available in the period prior to entry in the MHNIP, making comparison between treatment strategies (MHNIP and no MHNIP) difficult.

For those patients with a HoNOS score on entry to the MHNIP, and at one year (n=87), there was a statistically significant decrease in their HoNOS score (mean = -3.55; 95% CI -4.73 – -2.36; p<0.001). However, caution should be used when interpreting this change, as the mean HoNOS score for patients in this sub-group (mean = 13.69; 95% CI 13.00 – 14.38) was statistically different from the patients in our sample (mean = 15.56; 95% CI 14.86 – 16.26) (P<0.01).
Top of page

Discussion

MHNIP has the potential to reduce mental health related hospital admissions by approximately 3 days (95% CI -5.57 – 0.078) per patient with severe mental illness and would be associated with a cost saving per patient of around $2,600 (95% CI -$5353 – $75). This finding was statistically significant at the 0.10 level (p=0.058). These estimated savings might be conservative, given that additional savings may also be derived from a changing pattern of claims for MBS item numbers, and reduced attendances to emergency departments. This underestimate of resource utilisation may be particularly true for those patients whose illness may have been managed well, but then deteriorated rapidly (and therefore become 'at risk') in a short space of time before entering MHNIP. In this case, hospitalisation patterns in the 12 months prior to entry in the MHNIP may not representative of potential hospitalisations after entry into the MHNIP.

While the estimated savings of acute care spending ($2,600; 95% CI -$5353 – $75) for these patients is less than the indicative cost of providing the MHNIP service to these patients (metropolitan – $2,674; non metropolitan - $3,343) patient outcomes have also improved greatly. Patients and carers spoken to at case study sites overwhelmingly reported that MHNIP had assisted them in staying out of hospital and had helped them feel well and connected with their community. Qualitative information on patient outcomes, as reported by the HoNOS, have also improved during the first 12 months of receiving services under the MHNIP with an average aggregate HoNOS score reduction of 3.55 (95% CI -4.73 – -2.36).

An economic analysis incorporating a measure of patient utility was out of scope for this paper.

Table 8.5: Number of hospitalisations and length of stay, 12 months prior to entering MHNIP and 12 months after entering MHNIP by Primary Mental Health Diagnosis

Table 8.5 is separated into 4 smaller tables in this HTML version for accessibility reasons. It is presented as one table in the PDF version.

Number of patients

Number of patients for the following primary mental health diagnosis:
  • Adjustment disorder - 5
  • Anxiety - 27
  • Bipolar disorder - 33
  • Borderline personality disorder - 3
  • Depression - 137
  • Dysthymia - 6
  • Eating disorders - 2
  • OCD - 2
  • Organic personality disorder - 1
  • Personality disorder - 5
  • Post natal depression - 1
  • Post traumatic stress disorder - 4
  • Schizoaffective disorder - 2
  • Schizophrenia - 36
  • Unknown - 3
  • Total - 267

Period of 12 months prior to joining MHNIP

Primary mental health diagnosis# Hosp# Patients hospTotal LOSMeanSt Dev.95% CI Lower95% CI Upper
Adjustment Disorder
0
0
0
Anxiety
3
3
98
3.63
10.71
-0.41
7.67
Bipolar Disorder
6
6
87
2.64
6.52
0.41
4.86
Borderline Personality Disorder
0
0
0
Depression
12
11
297
2.17
7.83
0.86
3.48
Dysthymia
0
0
0
Eating Disorders
0
0
0
OCD
0
0
0
Organic Personality Disorder
0
0
0
Personality disorder
1
1
16
3.20
7.16
-3.07
9.47
Post Natal Depression
0
0
0
Post Traumatic Stress Disorder
6
1
12
3.00
6.00
-2.88
8.88
Schizoaffective Disorder
0
0
0
Schizophrenia
6
6
756
21.00
52.67
3.79
38.21
Unknown
0
0
0
Total
34
28
1266
4.74
21.35
2.18
7.30
Top of page

Period of 12 months after joining MHNIP

Primary mental health diagnosis# Hosp# Patients hospTotal LOSMeanSt Dev.95% CI Lower95% CI Upper
Adjustment Disorder
0
0
0
Anxiety
4
3
57
2.11
7.43
-0.69
4.91
Bipolar Disorder
6
4
195
5.91
24.73
-2.53
14.34
Borderline Personality Disorder
1
1
10
3.33
5.77
-3.20
9.87
Depression
14
9
231
1.69
7.04
0.51
2.87
Dysthymia
0
0
0
Eating Disorders
0
0
0
OCD
0
0
0
Organic Personality Disorder
0
0
0
Personality disorder
0
0
0
Post Natal Depression
0
0
0
Post Traumatic Stress Disorder
4
1
18
4.50
9.00
-4.32
13.32
Schizoaffective Disorder
0
0
0
Schizophrenia
1
1
21
0.58
3.50
-0.56
1.73
Unknown
0
0
0
Total
30
19
532
1.99
10.48
0.74
3.25

Paired t-test value (p)

Paired t-test values for the following primary mental health diagnosis:
  • Anxiety - 0.696
  • Bipolar disorder - 0.386
  • Borderline personality disorder - 0.42
  • Depression - 0.514
  • Personality disorder - 0.37
  • Post traumatic stress disorder - 0.39
  • Schizophrenia - 0.027
  • Total - 0.058

Footnotes

i In economics, cost-benefit analysis is used as a technical term to describe an analysis that provides information on the absolute benefits of one program or intervention over another. It requires all costs and benefits to be measured and reported in monetary terms. The theoretical properties of cost benefit analysis make this form of study highly attractive conceptually. In practice it is very difficult to implement comprehensively and is therefore rarely used for health sector evaluations outside the academic literature.

ii Andrews, G., et al., Tolkien II: A needs-based, costed, stepped-care model for mental health services: recommendations, executive summaries, clinical pathways, treatment flowcharts, costing structures.

iii Australian Department of Health and Ageing, 2011, Version 6 Final Service Weights, Cost Weights for AR-DRG Version 6.0x, Round 14 (2009-10) (www.health.gov.au/internet/main/publishing.nsf/Content/Round_14-cost-reports), retrieved September 2012.

iv The following DRGs were used to compute the 'weighted average per diem cost': U61A, U61B, U62A, U62B, U63A, U63B, U64Z, U65Z, U66Z and U67Z.

Top of page

prev page| TOC |next page