In my last column, I discussed evidence on how payment by results is actually working.

It confirmed that it has the potential to improve efficiency, while also debunking some of the urban myths that surround it - for example, that it has been draining resources from primary care, driving increased hospitalisations or undermining quality of care.

Equally, however, it is obvious that payment by results needs to improve further if it is going to deliver on its full potential. So here we consider some of the changes that should now be pursued, drawing on experience from other countries.

First a caveat. We should get away from the idea that payment by results can perform all the tasks required of commissioners. It can't. If we want fine grained and regularly evolving quality-related incentive payments to hospitals, in many cases they will work best if they are locally agreed payment overlays on the national payment by results system.

And if we want to tackle the£1bn of hospital admissions each year that could have been avoided by better ambulatory and primary care, primary care trusts will have to put in place alternative services and proper demand management tools. In short, the payment system can help with what economists call "technical efficiency" (improving the quality-adjusted cost of care), but "allocative efficiency" (getting the right mix of services) is mostly a job for the commissioners.

So where next for payment by results? One widespread idea is that we need more unbundling, in which more of the cost components of a unit of care are separable. Here, however, are four instances where what we need is more aggregation in the system, not less:

To incentivise hospitals to take patient safety seriously The tariff system does the reverse by paying for medical errors and then the extra services needed to remedy them. So like Medicare in the US, NHS commissioners should now cease to pay for hospital treatments arising from preventable errors. Payment systems should automatically edit out payments that relate to preventable events such as objects left in the body during surgery, air embolisms, blood incompatibility, pressure ulcers, patient falls, urinary tract infections and vascular catheter-associated infections.

In time it should be possible to go further, perhaps following the example of the Geisinger Medical Center in Pennsylvania, which now provides a 90-day guarantee on the cost of complications after heart surgery.

To offset the incentive for patients to be discharged prematurely When payment by results was devised, an incentive mechanism to prevent hospitals from discharging patients prematurely was built in, by proposing that they not be paid again for emergency readmissions within 28 days. It is, therefore, disturbing that the Audit Commission reports that PCTs have agreed to take this requirement out of their contracts, despite the fact that 83 per cent of them have seen increased readmissions. This provision needs to be reinstated.

To ensure neutrality of payment between site of care So, for example, by paying for inpatient and day-case elective surgery at a blended average, the tariff already incentivises efficiency improvement. And by paying for home births and hospital deliveries at a blended rate, the tariff will support mothers' childbirth choices.

To minimise complexity in the healthcare resource group system A payment system that participants struggle to understand is a system that will not produce desired behavioural change. The move to HRG4 is regrettably causing an explosion in the number of prices from 650 to 1,400. Given that the Australians manage on 665 codes and that Medicare, the world's largest taxpayer-funded tariff system, manages with 745 severity-adjusted diagnostic resource groups, there does seem to be an opportunity to simplify and streamline the classification system.

There are, however, other aspects of the tariff system where greater unbundling should indeed be permitted, starting with getting serious about the two thirds of mental health services still commissioned on antiquated pre-1991 block contracts. Rehabilitation services after acute admission are also fair game for this approach and HRG4s are supposed to allow this.

It also probably now makes sense to think about reimbursing email consultations, telephone advice and group visits, if the loss of old-style outpatient revenue is not going to stand in the way of these new patient-friendly modes of contact.

More controversially, if the tariff payment were to be split between the clinician and the hospital facility component, it would be possible to replace the consultant contract in at least some of the procedure-based specialties with a payment system for hospital doctors more closely linked to their individual and group productivity. That in turn might also make it easier to allow combined primary and secondary care teams to establish new community-based clinical models, immune from the gravitational economic pull of the general hospital.

This, after all, is how consultants are paid when engaged in private practice and well established systems exist internationally for doing so, such as the resource based relative value scale used by Medicare.

But perhaps the single most radical shift would be to move to a system of patient based, risk-adjusted commissioning "tariffs" on the PCT side of the system, to match those on the provider side. By this I mean that rather than setting PCT budgets by the old geographically based weighted capitation formula, we move instead to build up individual patient-based "dowries", the sum a commissioner would receive from the Department of Health for managing the total healthcare costs for that person for a year. This would unleash many possibilities for care delivery, particularly for people with long-term conditions, and is a proposal I will explain further in my column next month.