Right anterior thoracotomy aortic valve replacement is associated with less cost than sternotomy-based approaches: a multi-institution analysis of ‘real world’ data

Abstract

Background:

Large institutional analyses demonstrating outcomes of right anterior mini-thoracotomy (RAT) for isolated aortic valve replacement (isoAVR) do not exist. In this study, a group of cardiac surgeons who routinely perform minimally invasive isoAVR analyzed a cross-section of US hospital records in order to analyze outcomes of RAT as compared to sternotomy.

Methods:

The Premier database was queried from 2007–2011 for clinical and cost data for patients undergoing isoAVR. This de-identified database contains billing, hospital cost, and coding data from >600 US facilities with information from >25 million inpatient discharges. Expert rules were developed to identify patients with RAT and those with any sternal incision (aStern). Propensity matching created groups adjusted for patient differences. The impact of surgical approach on outcomes and costs was modeled using regression analysis and, where indicated, adjusting for hospital size and geographical differences.

Results:

AVR was performed in 27,051 patients. Analysis identified isoAVR by RAT (n = 1572) and by aStern (n = 3962). Propensity matching created two groups of 921 patients. RAT was more likely performed in southern hospitals (63% vs 36%; p < 0.01), teaching hospitals (66% vs 58%; p < 0.01) and larger hospitals (47% vs 30%; p < 0.01). There was significantly less blood product cost associated with RAT ($1381 vs $1912; p < 0.001). After adjusting for hospital differences, RAT was associated with lower cost than aStern ($38,769 vs $42,656; p < 0.01).

Conclusions:

Outcomes analyses can be performed from hospital administrative collective databases. This real world analysis demonstrates comparable outcomes and less cost and ICU time with RAT for AVR.

Keywords::

• Aortic valve
• Minimally invasive surgery
• Cost savings
• Outcomes

Introduction

The first valvular procedure via a right thoracotomy was reported by Lillehei et al.1 in 1957 when they performed an annuloplasty to correct mitral insufficiency. However, a full sternotomy became the preferred incision for valvular surgery for decades to come. It was not until 1993 that the first aortic valve replacement via a right anterior thoracotomy (RAT) approach was performed2. While several national champions of RAT for isolated aortic valve replacement (isoAVR) have presented their outcomes, no large multi-institutional analyses exist. Isolated series have exhibited equivalent results to sternotomy isoAVR while demonstrating less blood transfusion and decreased hospitalizations for RAT groups3–5. However, a major criticism of the RAT approach is that it increases the procedural costs due to the added equipment and longer operating times6,7. To this point, a group of experienced cardiac surgeons who routinely perform alternative incision AVR (Supplementary Appendix A) analyzed a cross-section of US hospital records in order to better understand ‘real world’ clinical and economic outcomes of RAT as compared to sternotomy approaches for isoAVR.

Materials and methods

Data source

The Premier hospital database, a highly regarded hospital database that is used extensively and cited regularly in these types of analyses, was used as the data source for this study8. This database contains complete patient billing, hospital cost, and coding histories from more than 600 healthcare facilities throughout the US. The data from which this study was derived was extracted from more than 25 million inpatient discharges and 175 million hospital outpatient visits from acute care facilities, ambulatory surgery centers, and clinics across the nation.

A protocol describing the analysis objectives, criteria for patient selection, data elements of interest, and statistical methods was submitted to the New England Institutional Review Board (NEIRB) and exemption was obtained (NEIRB # 13–203).

Eligible patients were 18 years of age or older and had undergone either a right anterior mini-thoracotomy (RAT) or any sternal incision (aStern) during the time period 2007–2011, for which isolated aortic valve replacement (isoAVR) was the primary reason for surgery. IsoAVR was identified for visits having the following primary ICD-9 procedure code classification: 35.21 and 35.22. Patients having either a record of Coronary Artery Bypass Grafting (CABG) in conjunction with isoAVR or procedures utilizing robotic technology were excluded. A set of expert rules were developed by authors ER, SCM, JRM, AS, GA, and EAG based on surgical experience to text mine the charge master billing files of the Premier database in order to identify a surgical approach as either RAT or aStern. Figure 1 displays rules for procedure identification and attrition.

Figure 1. Attrition diagram.

For all eligible patients, elements describing adverse events, hospital cost, surgery time, length of stay and re-admissions were obtained from the data. Cost analysis reflected the cost of the procedure to the hospital. The pre-operative All Patient Refined Diagnosis Related Groups (APR-DRG) severity level was used as an index of comorbidity. The 3 M APR DRG Classification System is a widely adopted proprietary risk adjustment classification tool, which uses information from routine claims data to produce valid and reliable severity measurement and risk adjustment scores9. It is used to account for differences related to an individual’s severity-of-illness or risk-of-mortality in large datasets. Comorbid conditions that might influence procedure selection or outcomes of interest, such as previous organ transplantation, the presence of pulmonary disease or diabetes mellitus, were obtained using ICD-9 diagnosis and procedure codes. Supplementary Appendix B provides a detailed listing of all ICD-9 codes for each condition included in the study. Information on socio-demographic characteristics and health insurance status was also included, as were descriptors of the care setting, namely census region, urban or rural setting, teaching hospital status, and facility bed count. Adverse events (identified by ICD-9 codes) that occurred intra-operatively were flagged and included in the analysis. A detailed list of each event and the corresponding ICD-9 code is found in Supplementary Appendix C.

Statistical analyses

The objective of this study was to use the Premier hospital database to compare clinical and economic outcomes in patients undergoing isoAVR with RAT vs aStern. Outcomes of interest included adverse events, hospital costs, length of stay, and re-admission rates. Propensity scoring was utilized to create well-matched groups for comparison. Comorbid conditions before and after matching are contained in Supplemental Tables 1 & 2. Since hospital costs have been found to be positively skewed, a generalized linear model (GLM) with a gamma distribution and a log link function was used to adjust for differences in hospital characteristics (teaching vs non-teaching, urban vs rural, and bed count) and to calculate the corresponding least square means of overall cost for the hospital stay within each matched cohort. Extensive details of the statistical methodology are contained in Supplementary Appendix D.

Results

Data analysis identified 27,501 patients undergoing AVR. Expert rule analysis positively identified isoAVR by RAT in 1572 patients and by aStern in 3962 patients. Propensity matching created two balanced groups of 921 patients each (Tables 1 & 2). RAT was more likely performed in southern hospitals (63% vs 36%; p < 0.01), teaching hospitals (66% vs 58%; p < 0.01), and larger hospitals (47% vs 30%; p < 0.01) (Tables 3 & 4). There were no statistically significant differences in adverse events between the two groups (Table 5). After adjusting for hospital differences, hospital lengths of stay and re-admission rates, up to 30–60 days, were not different between groups (Table 6). Cost analysis revealed that RAT was associated with lower hospitalization cost than aStern (Table 7) ($38,769 vs $42,656; p < 0.0001). Further examination revealed that RAT was associated with lower mean room and board costs, including ICU, compared to aStern ($11,283.56 vs $9890.77, p < 0.01). Finally, mean blood bank associated costs were lower for RAT group than aStern group ($1381 vs $1912, p < 0.001).

Table 1. Patient demographics before matching.

	RAT		aStern
Category	n	%	n	%
Total n	1572	100%	3962	100%
Age
<18	4	0.3	26	0.7
18–29	15	1.0	69	1.7
30–39	44	2.8	135	3.4
40–49	103	6.6	285	7.2
50–59	214	13.6	675	17.0
60–69	328	20.9	956	24.1
70–79	534	34.0	1181	29.8
80 or older	330	21.0	635	16.0
Race
Caucasian	1194	76.0	2889	72.9
African American	75	4.8	222	5.6
Hispanic	19	1.2	153	3.9
Other	284	18.1	698	17.6
Gender
Female	614	39.1	1635	41.3
Male	958	60.9	2327	58.7
Marital status
Married	902	57.4	2129	53.7
Unmarried	585	37.2	1666	42.1
Other/unknown	85	5.4	167	4.2
Insurance
Commercial	37	2.4	204	5.2
Medicare	963	61.3	2265	57.2
Medicaid	47	3.0	206	5.2
Managed care	482	30.7	946	23.9
Other	43	2.7	341	8.6
Health status
APR-DRG Severity  Level (1, 2)	646	41.1	1364	34.4
APR-DRG Severity  Level (3, 4)	926	58.9	2598	65.6

Table 2. Patient demographics after matching.

	RAT		aStern
Category	n	%	n	%	p Value
Total n	921	100%	921	100%
Age					0.9415
18–29	6	0.7	3	0.3
30–39	22	2.4	24	2.6
40–49	44	4.8	47	5.1
50–59	122	13.3	113	12.3
60–69	199	21.6	208	22.6
70–79	333	36.2	330	35.8
80 or older	195	21.2	196	21.3
Race					0.8767
Caucasian	785	85.2	795	86.3
African American	25	2.7	24	2.6
Hispanic	4	0.4	5	0.5
Other	107	11.6	97	10.5
Gender					1.0000
Female	367	39.9	367	39.9
Male	554	60.2	554	60.2
Marital status					0.7420
Married	578	62.8	582	63.2
Unmarried	319	34.6	320	34.7
Other/unknown	24	2.6	19	2.1
Insurance					0.9460
Commercial	13	1.4	10	1.1
Medicare	615	66.8	616	66.9
Medicaid	11	1.2	9	1.0
Managed care	269	29.2	271	29.4
Other	13	1.4	15	1.6
Health status					0.8501
APR-DRG Severity  Level (1, 2)	386	41.9	382	41.5
APR-DRG Severity  Level (3, 4)	535	58.1	539	58.5

Table 3. Hospital characteristics before matching.

	RAT				aStern
	By visit		By hospital		By visit		By hospital
Category	n	%	n	%	n	%	n	%
Total n	1572	100%	63	100%	3962	100%	92	100%
Census region
Northeast	360	22.9	13	20.6	732	18.5	13	14.1
Midwest	96	6.1	12	19.1	456	11.5	16	17.4
South	1058	67.3	28	44.4	1463	36.9	40	43.5
West	58	3.7	10	15.9	1311	33.1	23	25.0
Location
Urban	1508	95.9	60	95.2	3758	94.9	83	90.2
Not urban	64	4.1	3	4.8	204	5.2	9	9.8
Type
Teaching	1040	66.2	31	49.2	2204	55.6	39	42.4
Non-teaching	532	33.8	32	50.8	1758	44.4	53	57.6
Bed count
< 200	0	0.0	0	0.0	121	3.1	5	5.4
200–400	405	25.8	25	39.7	1127	28.5	41	44.6
401–600	383	24.4	19	30.2	1563	39.5	28	30.4
> 600	784	49.9	19	30.2	1151	29.1	18	19.6

Table 4. Hospital characteristics after matching.

	RAT				aStern
	By visits		By hospital		By visits		By hospital
Category	n	%	n	%	n	%	n	%	p Value
Total n	921	100%	59	100%	921	100%	71	100%
Census region									< 0.0001
Northeast	254	27.6	12	20.3	218	23.7	11	15.5
Midwest	52	5.7	12	20.3	117	12.7	11	15.5
South	584	63.4	27	45.8	331	35.9	32	45.1
West	31	3.4	8	13.6	255	27.7	17	23.9
Location									0.3872
Urban	880	95.6	56	94.9	872	94.7	65	91.6
Not urban	41	4.5	3	5.1	49	5.3	6	8.5
Type									0.0011
Teaching	604	65.6	28	47.5	536	58.2	31	43.7
Non-teaching	317	34.4	31	52.5	385	41.8	40	56.3
Bed count									< 0.0001
<200	0	0.0	0	0.0	42	4.6	3	4.2
200–400	235	25.5	24	40.7	292	31.7	29	40.9
401–600	255	27.7	18	30.5	313	34.0	23	32.4
>600	431	46.8	17	28.8	274	29.8	16	22.5

Table 5. Adverse events after matching.

	aStern (n = 921) n (%)	RMT (n = 921) n (%)	p Value
Neurologic complications^a	27 (2.9)	24 (2.6)	0.6701
Total Pulmonary Complications, (infections + non-infections)^b	312 (33.9)	283 (30.7)	0.1485
Pulmonary Complications, excluding
Infections^c	251 (27.3)	232 (25.2)	0.3142
Total infections^d	90 (9.8)	75 (8.1)	0.2210
Pulmonary infections^e	61 (6.6)	51 (5.5)	0.3296
Sepsis and Other Infections^f	40 (4.3)	32 (3.5)	0.3362
Wound Complications^g	55 (6.0)	49 (5.3)	0.5447
Acute Renal Failure	98 (10.6)	91 (9.9)	0.5909
Pacemaker Placement	69 (7.5)	65 (7.1)	0.7197
Atrial Fibrillation	386 (41.9)	404 (43.9)	0.3968
Re-admissions^h	319 (34.6)	346 (37.6)	0.1903
In-hospital deaths	17 (1.9)	25 (2.7)	0.2118

^aIschemic stroke, Hemorrhagic stroke or Transient cerebral ischemia/attack (TIA).

^bAcute respiratory failure, pulmonary insufficiency, atelectasis/pulmonary collapse, pneumonia.

^cLung comp-non-infectious.

^dPneumonia, Sepsis, bacteremia/SIRS, post-operative infection.

^ePneumonia.

^fSepsis, bacteremia/SIRS, post-operative infection.

^gHematoma/seroma/hemorrhage complicating a procedure, wound disruption/dehiscence.

^hAdmission to the same hospital within 60 days post-operatively.

Table 6. Healthcare utilization and costs after matching.

		Aortic
Category	Statistics	RAT	aStern
Total		921	921
Length of stay	median	7	7
	mean	8.73	8.78
	SD	9.2	6.44
Total hospital charges	median	$93,959	$124,054
	mean	$121,678	$153,427
	SD	$97,363	$98,184
Total hospital costs	median	$31,950	$38,721
	mean	$38,913	$45,607
	SD	$27,267	$25,924
Hospital costs
Blood Bank cost*	median	$670	$1,090
	mean	$1,381	$1,912
	SD	$2,476	$2,516
Cardiology/EKG	median	$625	$567
	mean	$1,144	$1,024
	SD	$1,349	$1,207
ER	median	$347	$324
	mean	$349	$343
	SD	$133	$148
Lab	median	$1,718	$2,478
	mean	$2,610	$3,454
	SD	$3,289	$3,344
OR	median	$7,751	$7,828
	mean	$8,106	$8,724
	SD	$4,587	$4,800
Other	median	$455	$297
	mean	$2,267	$1,955
	SD	$4,802	$7,114
Pharmacy	median	$1,965	$2,237
	mean	$3,103	$3,744
	SD	$5,063	$7,610
Radiology	median	$436	$474
	mean	$687	$951
	SD	$958	$2,106
Respiratory	median	$636	$920
	mean	$1,146	$1,409
	SD	$1,957	$1,915
Room and board*	median	$6,207	$8,257
	mean	$8,731	$10,535
	SD	$12,919	$9,319
Room and board (ICU)*	median	$2,982	$3,777
	mean	$5,123	$5,666
	SD	$8,808	$6,612
Supply	median	$9,762	$11,709
	mean	$11,054	$13,622
	SD	$5,706	$8,407
Therapy	median	$274	$317
	mean	$418	$485
	SD	$579	$528

*p < 0.001.

Table 7. Least square means for total hospital costs.

By surgery type:
Effect	Estimate	p Value
aStern	$42,655.97	< 0.0001
RAT	$38,768.85

LS means adjusted for region, teaching status, urban status and number of beds. Total costs were modeled using generalized linear models with a gamma distribution and log link due to the inherent distribution of cost data.

Discussion

This study used a conservative ‘real world’ analysis (grouping full and partial sternotomies) and demonstrated comparable outcomes as well as less total cost, room and board cost, as well as blood bank related cost with RAT for isoAVR. While others have published single center series comparing minimally invasive AVR (RAT or hemi-sternotomy vs sternotomy), this is the first large multi-center/multi-regional study comparing RAT to any sternotomy approach for isolated AVR.

After propensity matching, we did not appreciate significant differences in mortality or complication rates between groups. This study corroborates the findings of several other studies demonstrating that RAT for AVR is safe, with complications rates similar to those using a sternotomy approach3,4,10,11. Therefore, based on this study as well as many other single center studies, RAT for isolated AVR appears to be safe.

Multiple series have demonstrated shorter hospital LOS when a less invasive approach (RAT or hemi-sternotomy) is used instead of a full sternotomy3–5,12–17. Our study did not show a difference in hospital LOS. A potential explanation is that the aStern group included both hemi- and full sternotomy patients and, thus, it could have benefited the LOS for the aStern group as hemi-sternotomy has been shown to be associated with shorter LOS compared to full sternotomy15–17. ICU LOS has also been shown to be shorter when less invasive techniques, such as hemi-sternotomy, are used for AVR15,16. We were not able to obtain specific ICU LOS vs total hospital LOS due to limitations in the Premier database. Nonetheless, the mean ICU room and board cost was higher for the aStern group compared to RAT group ($5666 vs $5123). This may suggest a shorter ICU LOS or lower costs due to decrease intensive care needed for the RAT group.

Our study demonstrated roughly a $500 (28%) reduction in blood bank related costs for RAT patients despite there being a few more patients in the RAT group receiving transfusion (n = 871; Table 6). Our analysis confirms less blood product usage for RAT as suggested by other series3.

Finally, this study also demonstrates the potential value of developing prospective query methodologies in order to probe large hospital administrative collective databases, which comprise broad geographic and demographic representation of US hospitals, in order to compare different surgical approaches. This approach allows investigators to effectively compare outcomes utilizing relatively recent data for large and broad patient populations representing the real world setting.

Limitations

This study also had some noteworthy limitations. The authors acknowledge that ‘minimally invasive’ cardiac surgery is often recognized as including both partial sternotomy and RAT approaches. Based on expert rules for procedure identification, it was not feasible to clearly segregate partial from full sternotomies and, therefore, both are represented in the aStern cohort. Additionally, due to the de-identified nature of Premier, an audit of an institution’s data to verify the expert rules algorithm was not possible. However, this large multi-institutional comparison of sternal vs non-sternal approach for isoAVR is compelling, filling a gap in the literature. Furthermore, given the demonstrated clinical benefits of partial-sternotomy vs full sternotomy16,18,19, it could be argued that the observed differences between the two cohorts may have been even greater if the partial-sternotomies had been excluded from the control group.

While the data were mined from a sophisticated hospital administrative database, certain data points could not be clearly identified. These limitations are inherent to the data source; however, they could be rationalized to impact both cohorts similarly. Additionally, the analysis was limited to a 30–60 day perioperative period, which restricts analysis of potential long-term complications.

Conclusion

Informative outcomes analyses can be performed from hospital administrative collective databases. This conservative real world analysis (grouping full and partial sternotomies together) demonstrates mostly comparable clinical outcomes but lower costs associated with right anterior mini-thoracotomy for AVR.

Transparency

Declaration of funding

Funding for this project was provided by Edwards Lifesciences. This included funding for access to the Premier database and statistical analysis.

Declaration of financial/other relationships

ER, SCM, AS, GA, EAG are Edwards Lifesciences consultants. MM is an Edwards Lifesciences employee. JRM is on the Edwards Lifesciences medical advisory board. CG is an employee of CTI Clinical Trials and Consulting services which is a paid consultant to Edwards Lifesciences. ER is also a speaker for Medtronic Sorin and is on the Cardionet advisory board. SCM is a Bazter consultant and an Abiomed speaker. GA is an Abbott and Mitralign consultant and an Amcure and St Judes speaker. EAG is a Medtronic consultant and holds Intellectual property with Edwards Lifesciences and Medtronic. JME peer reviewers on this manuscript have no relevant financial or other relationships to disclose.